PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261001 (2023) https://doi.org/10.1117/12.2682191
This PDF file contains the front matter associated with SPIE Proceedings Volume 12610, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261002 (2023) https://doi.org/10.1117/12.2671394
The calculation of transcendental functions is a common step in industrial control algorithms. Taking up a large number of CPU cycles, the transcendental function calculation by software compresses the computing resources of real-time control algorithms. As the complexity of industrial control system increases, hardware accelerators have become one of the strategies to solve this contradiction. In this paper, we designed a multi-thread, high-performance, configurable hardware accelerator for transcendental functions, which is based on the iterative calculation. It supports the calculations of sine cosine, arc-tangent, modulus, exponent and logarithm. The accelerator uses the standard cell library of SMIC 40nm Eflash platform for synthesis, achieving a frequency of 200MHz and a synthesis area of 176,840um2.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qiang Li, Zhu Liu, Hengtao Liu, Lei Yang, Chunyang Li
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261003 (2023) https://doi.org/10.1117/12.2671446
With the vigorous development of electric vehicles and the arrival of the era of big data, electric vehicles and related data are becoming more and more perfect. At present, the research method of electric vehicle charging demand prediction is mainly based on the travel chain, traffic network, combined with users' charging choice behavior and other factors to simulate their charging demand. How to accurately describe the travel chain, traffic network and other factors is the focus and difficulty of the research. In the context of the Internet, it has contributed to the demand for charging at any time. In order to meet this demand, the establishment of a new charging distribution service mode has improved people's charging methods, but the resulting number of Charging Station (CS) distribution orders has increased significantly. The pressure of production and distribution of CSs has risen steadily, and crowdsourcing operation and maintenance mode came into being. The decision planning and intervention measures to encourage order receiving are studied respectively for the two cases of multi person order receiving and no one order receiving in crowdsourcing operation and maintenance distribution. Under the condition of multiple people grabbing orders, a comprehensive decision-making model based on Computer-Aided Technology (CAT) is established, and an example of the decision-making model is analyzed to verify the scientific value and effectiveness of the model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261004 (2023) https://doi.org/10.1117/12.2672640
Aiming at the problem of tracking and controlling the motion path of industrial robots in the process of research, design and development, this paper will take the common six-axis industrial robots as the research object, take advantage of the application advantages of VR technology, 3D modeling technology and Web3D interactive technology, take 3ds Max as the modeling tool and Unity3D virtual reality engine as the development platform, and build a virtual reality simulation experiment system of industrial robots from the perspective of visual interaction between virtual robots and real robots, so as to provide a comprehensive and feasible solution for the research of virtual motion simulation and control of industrial robots. The whole system adopts B/S architecture and completes the design and deployment of the whole function according to MVC mode in APS.NET environment, so as to support users with different roles to test the functions of each component module of industrial robot in virtual reality environment, and also simulate the trajectory planning and motion effect control of industrial robot in different scenes. The system will greatly improve the research and development efficiency of industrial robots, increase the efficiency and flexibility of industrial robots, break through the limitations of traditional testing methods on time and space, and provide experience and reference for the intelligent development of industrial robots.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261005 (2023) https://doi.org/10.1117/12.2671565
New energy has penetrated power grid companies on a large scale, affecting the cost of power grid planning, operation scheduling, overhaul, and maintenance, etc. However, the impact of different types of new energy power sources on power grid costs is complex and diverse, making it difficult to accurately quantify the impact of new energy penetration on the entire cost chain of the power grid. For this reason, based on system dynamics, this study constructs a full cost chain model of the power grid under the penetration of new energy, and quantitatively analyses the impact of penetration of new energy on the full cost of the power grid under different scenarios. Finally, the correctness of the model is verified through case analysis, and the optimal penetration under different new energy penetration scenarios is obtained, which provides managers with a basis for decision-making, and ultimately maximize the benefits of power grid companies.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261006 (2023) https://doi.org/10.1117/12.2671254
Aiming at the problem of low model recognition accuracy caused by high similarity and mutual occlusion between crops and weeds in Unmanned Aerial Vehicle (UAV) images, a pixel-level weed recognition method based on improved HRNet-OCRNet is proposed. In this method, a multi-stage and multi-scale feature fusion method is added to HRNet to preserve more details and enhance semantic information at different levels, to solve the problem of high similarity between crops and weeds. The spatial self-attention module of Polarized Self-Attention (PSA) is integrated to HRNet, enhance the network's learning of important features, and reduce the false identification caused by mutual occlusion of crops and weeds. The expansion prediction method is used to generate an accurate distribution map of crop weeds. Compared with Deeplabv3+, GCNet and K-Net, the experimental results show that the proposed method has higher recognition accuracy for crop weeds, and mean intersection over union (mIoU) reaches 85.76%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261007 (2023) https://doi.org/10.1117/12.2671511
For different application scenarios, aerospace vehicle control systems may use different flight control hardware platforms, flight control strategies, and control functions, and the variability in adoption may lead to difficulties in the development and transplantation of software for general purpose. In this paper, we propose a new idea to adopt modular design to decompose the vehicle control software functions into sensor data unpacking, navigation calculation, attitude control calculation, data fusion, control output, hardware driver, thread scheduling and other modules, modular packaging design or each function, and hierarchical design for the overall architecture to achieve the minimum modification to meet the development needs of different application scenarios. The overall architecture is designed in a hierarchical way to meet the development needs of different applications with minimal modifications and good scalability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261008 (2023) https://doi.org/10.1117/12.2671301
Due to installation technology, secondary span oscillation, galloping, icing and other factors in transmission lines, crimping pipes are prone to many defects. Because of the height problem, it isn't easy to detect it. Therefore, this paper designs a live detection method based on the Unmanned Aerial Vehicle (UAV) for the defects of transmission line crimping tubes. The quadrotor UAV intended in this paper comprises a flight control system, fuselage, power and energy devices, etc. It is used to carry imaging systems and detection equipment. The image processing software designed in this paper preprocesses the image to obtain a precise fault location. A digital imaging system of an X-ray machine with live working is created, which is mounted on the UAV to collect the images of the defects of power transmission line crimping tubes. The UAV is equipped with live detection equipment for crimping tube defects to realize the nondestructive live detection of defects such as scratches on the steel core surface of crimping tubes and broken steel core strands. The test results show that this method has a low error detection rate for both functional and technological defects and realizes comprehensive defect detection.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261009 (2023) https://doi.org/10.1117/12.2672792
In order to build an aeroengine on-board model with full envelope, full state, high accuracy and high real-time, a modeling method based on flight data is proposed. This method builds state variable model based on component level model. Considering the influence of Reynolds number, power extraction, air bleed and other factors, the steady state model of the on-board model is modified based on regression analysis using flight data to reduce the modeling error caused by individual engine differences. At the same time, in order to compensate the residual steady-state error, a steady-state error model based on Gaussian Mixture Model Neural Network (GMM-NN) is established. Considering the need to reconstruct the speed sensor, the speed signal cannot be used as the scheduling variable to build a new scheduling variable, which has less dynamic error compared to taking fuel as the scheduling variable. Compared with the traditional model, the input of this model is only control variables and flight conditions, and it can reconstruct the signals of speed, pressure, temperature and other sensors. At the same time, it has the advantages of simple structure, no iterative calculation and high accuracy. Compared with flight data, the maximum dynamic error of compressor outlet total pressure of the new scheduling variable model is 3.564%, which is 4.13 times higher than the maximum relative error of 14.735% of the fuel scheduling model. In the verification of multi flight data, the average errors of LP rotor speed, HP rotor speed, compressor outlet total pressure and LP turbine outlet total temperature are 0.52%, 0.39%, 0.53% and 0.9% respectively, meeting the accuracy requirements of the project.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100A (2023) https://doi.org/10.1117/12.2671248
Human-Object Interaction (HOI) detection is a fundamental task for understanding real-world scenes. In this paper, a graph model-based human-object interaction detection algorithm is proposed, which aims to make full use of the visual-spatial features and semantic information of human-object instances in the image, thereby improving the accuracy of interaction detection. Aiming at the characteristics of visual-spatial features and semantic information, we take the visual features of human and object instance boxes as nodes, and the corresponding spatial features of interaction relations as edges to construct an initial dense graph, and adaptively update the graph through the spatial and semantic information of instances. The V-COCO dataset is used to evaluate the algorithm, and the final accuracy is significantly improved, which proves the effectiveness of the algorithm.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100B (2023) https://doi.org/10.1117/12.2671168
Developing a safe, stable, and efficient obstacle avoidance policy in crowded and narrow scenarios for multiple robots is challenging. Most existing studies either use centralized control or need communication with other robots. In this paper, we propose a novel logarithmic map-based deep reinforcement learning method for obstacle avoidance in complex and communication-free multi-robot scenarios. In particular, our method converts laser information into a logarithmic map. As a step toward improving training speed and generalization performance, our policies will be trained in two specially designed multi-robot scenarios. Compared to other methods, the logarithmic map can represent obstacles more accurately and improve the success rate of obstacle avoidance. We finally evaluate our approach under a variety of simulation and real-world scenarios. The results show that our method provides a more stable and effective navigation solution for robots in complex multi-robot scenarios and pedestrian scenarios.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100C (2023) https://doi.org/10.1117/12.2671382
The surgical robot, a crucial instrument in the emerging field of Minimally Invasive Surgery (MIS), has clear advantages over traditional MIS in terms of flexibility, accuracy, and 3D vision. A crucial computer vision task that helps surgeons is the real-time and precise segmentation of the region of interest during robotic surgery. In this paper, a brand-new and reliable learning framework for image segmentation is proposed, and it is based on a single RGB camera. The major study goal is presented after a brief introduction to the current state of affairs and the visual capabilities of surgical robots at first. Then, a brief review and discussion of the existing research on deep learning-based computer vision in robotic surgery is given, followed by the presentation of our suggested deep neural network structure. Additionally, the comprehensive extraction of the multi-scale feature information makes use of the atrous convolutional procedure. Finally, a variety of image segmentation evaluation metrics are introduced, and the performance is promising for the clinical domain.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Peipei Xu, Lianxiang Jiang, Bingui Xu, Mingxiang Li, Fei Wang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100D (2023) https://doi.org/10.1117/12.2671425
Low-cost, intelligence and short development cycle has become its trend of small satellites. A hybrid on-board avionics topology based on CAN bus and router was proposed. The telemetry was collected by On-Board Computer (OBC) via CAN bus, while the router integrated RS422, LVDS, Ethernet, Camera Link and TLK2711 interfaces, which support data rate varying from 1Mbps to 10Gbps and usually used by payloads, so it makes regular payloads integrated into the avionics much easier. The OBC used the PowerPC MPC8548 processor, which run at 1GHz. Plug and play mechanism was adopted to make the OBC recognize the devices dynamically when they powered on, which accelerated the system integration; furthermore, the software modules were also allowed to install or uninstall dynamically on-line for flexibility. For the modular and various interfaces supported, payload modules such as GNSS-R receiver, ADS-B receiver and camera electronics was easily integrated into the avionics box, so the signaling were transferred via the backplane instead of cables.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100E (2023) https://doi.org/10.1117/12.2671398
In recent years, due to the development of Internet technology, more and more people begin to connect their daily life with internet technology. Due to its intelligence, simple operation and daily characteristics, wearable devices have gradually entered the public's vision and been accepted by the public. More and more wearable devices appear with a variety of supporting software. This thesis explores how this software use different middleware for different functions, design the UI interface to make people easy to use and use different ways to protect user privacy. To analyze the advantages and disadvantages of the existing software products of wearable devices and propose changes and improvements for future software development, firstly find out some existing wearable device software, and analyze its data processing, functions, adopted protocols, user privacy management, and other aspects. Secondly, through searching corresponding thesis and questionnaires, finds users' dissatisfaction with existing wearable devices. Finally, based on the feedback of users, the data processing, privacy protection and other functions of the existing wearable device software are summarized, and the corresponding direction for the future improvement of wearable devices is put forward.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100F (2023) https://doi.org/10.1117/12.2671237
To solve the problems of insufficient local search capability and easily falling into local optimization in the Aquila Optimizer (AO), an aquila optimizer integrating Gaussian walk and somersault strategy (AO-IGWSS) is proposed. Strengthening the exploitation ability, a Gaussian walk strategy is used instead of Levy flight to generate step size adaptively controlled by iteration numbers. Furthermore, to enhance the capability of local optima avoidance, a somersault strategy is introduced to update individuals. The experimental results on nine benchmark test functions prove that the AO-IGWSS can achieve better results than the original AO algorithm, the differential evolution mutation and tangent flight aquila optimizer (DEtanAO), and four other intelligent optimization algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100G (2023) https://doi.org/10.1117/12.2671083
Voice interactivity is one of the critical factors influencing user behaviors and intention in human-computer interactions, but little is known about how voice interactivity between smart voice robots (e.g., smart speakers) and users or among users affects user satisfaction in the robot-user interaction context. The authors employed an online survey to explore the influential mechanism between voice interactivity with or via smart voice robots and user satisfaction. The results showed that: (1) robot-user interactivity positively affects psychological ownership and flow experience; (2) user-user interactivity positively impacts psychological ownership and flow experience; (3) psychological ownership positively influences user satisfaction; (4) flow experience has a positive effect on user satisfaction. This study discussed the theoretical and practical implications of the above findings.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100H (2023) https://doi.org/10.1117/12.2671060
Due to the complex spatio-temporal correlation of meteorological data, weather forecasting is a challenging task. Recently, with plenty of meteorological data available and the successful applications of deep learning technology in many areas, developing data-driven models for this task has achieved great attention. Especially, Convolutional Recurrent Neural Networks (CRNNs) have been shown to be effective in spatio-temporal predictive learning. The convolutional connection with shared weights is fixed for different spatial locations and timestamps, while spatio-temporal transformations of meteorological data are varying in both time and space. To address this problem, we developed a Spatio-Temporal Adaptive Convolution for the Gated Recurrent Unit (GRU) to improve the ability of extracting spatio-temporal features. For convenience, we abbreviate our model as STAConvGRU for weather forecasting. The key motivation behind STAConvGRU is to develop additional convolution layers under the framework of the ordinary RNN to learn simultaneously the sampling positions and weights of convolutional kernels. As a result, the adaptive convolution could select the positions and adjust the weights according to the spatio-temporal information. Comparative experiments are conducted on four types of meteorological datasets, including temperature, relative humidity, wind, and radar echo. The experimental results demonstrate the effectiveness and superiority of our proposed model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100I (2023) https://doi.org/10.1117/12.2671250
Face recognition has been widely used in daily life, but the existing model systems use processed high-quality datasets in training, while the face pictures in real scenes usually contain the influence of blurring, lighting, obscuring and other factors, thus making the existing face recognition models cannot perform well, and secondly, the existing face datasets have less data of Asian descent, resulting in the distribution learned by the models with the actual application. There is a certain error in the actual application. We propose a method to train face recognition models for realistic scenes by image augment of local face data to improve the classification accuracy of the models for low-quality images, and we demonstrate the feasibility of our method through experiments. Our method improves 0.619% and 0.414% in classifying images with added illumination and added random squares, respectively, compared to the current state-of-the-art methods.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100J (2023) https://doi.org/10.1117/12.2671404
The traditional artificial potential field method, distance is the only factor to determine the potential field force. When the UAV enters the obstacle's range of action, it is repelled by its potential field, the obstacle will have a repulsive effect on the UAV and as the distance continues to approach, the UAV is subjected to more and more repulsive force, making the UAV avoidance time is too long and the avoidance path is wasted. This paper proposes an improved artificial potential field method for the UAV forward path and obstacles do not intersect and is still in the variety of action of the repulsive potential field, which solves the problem that when the UAV forward direction does not intersect with the obstacles, the UAV is in the range of action of the repulsive potential field and is not subject to repulsive force, avoiding the waste of obstacle avoidance path. It is demonstrated through simulation analysis that the proposed obstacle avoidance algorithm produces superior results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Yi Pan, Tiankui Sun, Xin Fang, Xiaodong Yuan, Mingming Shi, Jinggang Yang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100K (2023) https://doi.org/10.1117/12.2671179
Under the background of the modern energy systems, the degree of interconnection and coupling between the power system and natural gas systems is gradually increasing. The economic dispatch model is becoming more and more complex. However, the economic dispatch of an integrated energy system is a high-dimensional, multi-period optimization problem, especially the multi-period joint optimization will greatly reduce the convergence speed of the model. Moreover, the electric power system and the natural gas system belong to two independent systems, which have a limited understanding of each other's information changes. In this paper, the cost of comprehensive operation is considered, and then the objective function is modeled. Based on the idea of k-means clustering, a granularity partition algorithm, which is based on the improved k-means clustering, is designed. To improve this algorithm, this paper adopts the shortest distance method to cluster the original data. Then, it adopts the rough clustering results to calculate the clustering center, and then applies the k-means clustering. From the source side heating coupling optimization and the load side load curve, the introduction of an air source heat pump at the source side and the consideration of waste heat recovery of electricity to the gas device can improve the flexibility of energy supply mode and optimize the heat supply coupling at the source site. Besides, it can equivalently decouple the problem of wind curtailment caused by the constraint of "determining power by heat" in co-generation, realize electric energy substitution and reduce the carbon emission of co-generation. Based on the principle of calorific value equivalence, the load side analyzes the energy conversion value of electric and thermal users and forms a comprehensive demand response mechanism with energy substitution under the guidance of various energy price signals to assist wind power integration and consumption. Through the coordination and optimization of source and load, the phenomenon of wind power curtailment is significantly reduced while ensuring the economic operation of the system.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100L (2023) https://doi.org/10.1117/12.2671697
Due to the relative maturity of Data Flow (DF) analysis and research technology as well as the increasing number and development of diagnostic instruments, the automotive engine fault detection and diagnosis system is developing in the direction of precision, intelligence and multi-module diagnosis. In this paper, the fault detection system is designed for the problem of insufficient power of automobile engine. The ECU of automobile engine has the function of collecting the engine DF under various conditions in real time, and then reading out the relevant data during the driving process with the automobile fault diagnostic instrument, and the maintenance personnel combine them according to different requirements to form different data sets. The fault detection system also has the function of fault identification and prediction, and the best result of PSO-ELM algorithm is found by comparing the prediction accuracy of different algorithms. Here, it is hoped that the research in this paper will provide a reliable basis for analyzing and researching automotive engine faults.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Xu Zhao, Xuecheng Du, Xu Xiong, Zhi Chen, Wei Jiang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100M (2023) https://doi.org/10.1117/12.2671259
Field-Programmable Gate Array (FPGA) is now considered by an increasing number of designers in various fields of application, such as space and aircraft embedded control systems, image and signal processing. It is a popular way to design different serial communication protocols including serial and network communications implemented by FPGA, which ensure data transfer accuracy and timing. However, Single Event Effects (SEE) can pose a serious threat to the reliability of FPGA in the radiation environments. In order to evaluate the reliability and failure modes of different communication interfaces in radiation environments, three communication protocols were tested by alpha radiation source, including Universal Asynchronous Receiver Transmitter (UART), ethernet, and Universal Serial Bus (USB). The experiments have presented the failure modes and SEE cross sections to different communication ways.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100N (2023) https://doi.org/10.1117/12.2671375
Sandpaper is an indispensable and important article in industrial production and daily life. However, due to technological reasons, not all sandpaper can meet the standard. At present, in the traditional industrial production, in order to improve the quality of sandpaper, the usual way to identify the problem sandpaper is human eyes. Due to the influence of personnel factors, the accuracy and reliability of this way are low, and the production cost is high, and the efficiency is low. Therefore, we designed an automatic detection system of sandpaper defects based on digital image processing technology, which can effectively improve the recognition rate of defective sandpaper, reduce the production cost and improve the production efficiency.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100O (2023) https://doi.org/10.1117/12.2671323
At present, the data cleaning method based on time series realizes the data cleaning by classifying the data in the time series. Due to the lack of dimensionality reduction, the cleaning efficiency is low. For this reason, this paper proposes a method for rapid cleaning of smart grid data based on sparse self-coding. In this paper, the encoder neural network is constructed to reduce the dimension of the data, and Logsf algorithm is used to obtain the optimal weight of the data, obtain the main characteristics of the data, and achieve clustering cleaning of the data. In the experiment, the cleaning efficiency of the proposed method was verified. The experimental results show that the method proposed in this paper has a short time delay and high cleaning efficiency for smart grid data cleaning.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100P (2023) https://doi.org/10.1117/12.2671458
Medical robot is an intelligent service robot. It’s one of the most active and invested fields in the field of robot research abroad, and their development prospect is very promising. The question for this research is how to design the interaction process of intelligent medical robots in line with the concept of human-computer interaction. The purpose of this study is to summarize the interaction process and framework between man and machine, so as to be further applied in the future medical industry and provide patients with more trustworthy services. The robot is mainly divided into three types: monitoring, surgery and service, and the interaction process design of the three types of robots is analyzed respectively. This study will verify the application value of human-computer interaction from these three perspectives. It is of great significance for promoting the development of artificial intelligence in the medical field.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Lixue Tian, Xianjie Huo, Zhishuai Zhang, Wenzhe Ma, Yingtao Li
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100Q (2023) https://doi.org/10.1117/12.2671149
In this paper, low dropout voltage regulator based on Flipped Voltage Follower (FVF) is proposed. It is futured with fast transient response to load changes, high slew rate, and faster power-on time. The circuit proposed in the paper is adopted with double error amplifier and transient enhancement circuit. The major error amplifier can provide reference voltage for FVF structure and ensure flipped voltage stability. The auxiliary error amplifier is used to form a feedback loop at VOUT and VREF, improving the precision of output voltage, generating extra charge and discharge branch to provide greater slew rate, faster loop response, and faster power-on speed. The simulation results show that this LDO has output voltage jump less than 33mV and 150ns power-on time.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100R (2023) https://doi.org/10.1117/12.2671159
Diversity among the members of classifiers is deemed to be a key point in classifier ensemble. However, there doesn’t exist a widely accepted diversity measure and construct. In this paper, we propose a sample and feature double random construction of training sample variability. A support vector machine is used as the base classifier to construct the difference by distinguishing the regularization term C and the kernel function. Based on the negative correlation theory, the base classifier generalization error and disparity judgment functions are proposed, and the base classifier is integrated by ranking according to the judgment functions, which could achieve a higher accuracy rate by the support vector machine ensemble.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100S (2023) https://doi.org/10.1117/12.2671212
Lane detection is a crucial environmental sensing technique that is used in advanced driving assistance systems and automatic driving. The research on this issue has significant practical value. Aiming the current lane detection algorithm could not solve the problems of the local receptive field and detail feature loss, we introduced the multi-head self-attention module in Transformer into the encoder and decoder to obtain the global receptive field while solving the problem of detail feature loss with the multi-level feature fusion decoder. The proposed algorithm has been compared with the ERFNet model in the CULane dataset, and the detection accuracy has improved by 3.9 percentage points. The detection accuracy in the Tusimple dataset is 96.51%. Introducing a multi-head self-attention module increases the feature selection effect of the attention mechanism in the coding and decoding process. It provides a new solution for the lane detection algorithm.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100T (2023) https://doi.org/10.1117/12.2671471
In the process of driving a car, negative emotions such as anger, uneasiness, and nervousness of the driver seriously affect the personal safety of the driver and passengers. However, due to various reasons during the driving process, many factors will prevent the driver from showing true emotions on the face, which will seriously affect the accuracy of identification. And Real-time detection of physiological signals to detect emotions requires complex instruments, which is difficult to carry out in real driving. In order to solve this problem, the authors proposed a framework for driver emotion detection, the proposed algorithm includes three Models: (1) Automotive Advanced Driver Assistance System ADAS emotion monitoring framework. (2) Deep learning-based human facial expression recognition algorithm convolutional neural network CNN to perform human facial expression recognition. (3) Set a threshold window to decide whether to remind. After 100 epochs of training on the dataset of FER-2013, the model accuracy is 60.5%. The model can accurately classify facial expressions in most cases by volunteer testing.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100U (2023) https://doi.org/10.1117/12.2671052
In order to improve the control accuracy of aero-engine operation limits, a novel design approach of Min-Max limit protection is proposed. The inverse mapping models of different limits based on On-Line Sliding Window Deep Neural Network (OL SW DNN) are proposed and established firstly. The OL SW DNN models calculate the limit value of fuel flows to ensure that engine satisfies all operation limits. The operation restrictions in the proposed method can vary in different flight conditions. With the application of on-line learning modeling method, the engine can always operate within the given operation limits no matter whether engine degrades or not. Moreover, the OL SW DNN adopts deep learning structure and has strong fitting capacity for the nonlinear object. The comparison simulations of the popular limit protection design method based on optimization method and the proposed one are carried out. Compared with the popular method, the limit line of each operation limits in the proposed method not only has much higher accuracy especially when engine appears degradation, but also can be continuous change.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100V (2023) https://doi.org/10.1117/12.2671522
Due to the limitation of computer capacity and energy of equipment, unmanned equipment cannot perform intensive computer tasks well during emergency failure inspection. In order to solve the above problems, this paper proposes a task waste strategy based on Deep Reinforcement Learning (DRL), which is mainly applicable to several UAVs and individual ES scenarios. First of all, an end edge cloud cooperative unloading architecture is built in the edge environment of UAV, and the problem of unloading tasks is classified as an optimization problem to achieve the minimum delay under the limit of the computing and communication resources of the Edge Server (ES). Secondly, the problem is constructed as Markov decision, and Deep Q Network (DQN) is used to solve the optimization problem, and experience playback mechanism and greedy algorithm are introduced into the learning process. Experiments show that the mitigation strategy has lower latency and higher reliability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100W (2023) https://doi.org/10.1117/12.2671205
This paper proposes an end-to-end abnormal behavior detection network to detect strenuous movements in slow moving crowds, such as running, bicycling in transportation surveillance videos. The algorithm forms continuous video frames into a video packet and use the video packet feature extractor to obtain the spatio-temporal information. The implicit vector-based attention mechanism will work on the extracted video packet features to highlight the important features. We use fully connected layers to transform the space and reduce the computation. Finally, the packet-pooling maps the processed video packet features to the abnormal scores. The network input is flexible to cope with the form of video streams, and the network output is the abnormal score. The designed compound loss function will help the model improve the classification performance. This paper arranges several commonly used anomaly detection datasets and tests the algorithms on the integrated dataset. The experiment results show that the proposed algorithm has significant advantages in many objective metrics comparing with other anomaly detection algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100X (2023) https://doi.org/10.1117/12.2671053
The combination of low and high order features in the recommendation system is crucial to the predicted click through rate. This paper designs an Attention Deep Cross Attention Recognition Machine (ADCAFM). Traditional recommendation models only use attention factor decomposers and deep cross networks to extract low and high order features, but the diversity of deep cross networks mining user interests is weak. Therefore, this paper extracts the feature depth of different subspaces by integrating the multi head attention mechanism to solve the problem of user interest diversity in deep cross network mining; Finally, the low and high order combined features are effectively fused and recommended together. Through experimental comparison on Criteo and Movielens-100K data sets, the AUC index is used for evaluation. Compared with the benchmark model, the AUC index is 1.85% and 1.55% higher.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100Y (2023) https://doi.org/10.1117/12.2671177
This study proposes an innovate nonlinear programming model to optimize the shipping cost under the restriction of Emission Control Area (ECA). The arctic shipping route entitles the most potentially route from Asia to Europe to significantly reduce the shipping cost. Therefore, aiming to optimal the existing shipping cost, the optimized model is provided and finds that the feeder ships have better performance on economic cost whether sailing in ECA accordingly with the carbon tax less than 190 USD/tons. The proposed optimization method can well reduce the shipping cost and get better emission performance when choosing the arctic shipping route. And the results also can improve the shipping company’s revenues and maintain environmental sustainability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126100Z (2023) https://doi.org/10.1117/12.2671080
In order to improve the detection accuracy and efficiency of overprinting, an overprint deviation detection method is proposed for the characteristics of "concentric circles" overprint logo logos with fixed radius and line width. Firstly, the edges are divided into foreground and background by edge enhancement and center-of-mass comparison to select the appropriate high and low thresholds; the image edge information is extracted by Canny operator and the outline is filtered according to the radius of the circle in the logo logos; finally, the circle center is confirmed by the least squares method, and the distance between the two circle centers is the overprint deviation. deviation, and the pixel deviation is converted to physical deviation by the size of the camera calibration. Comparing the data obtained by this method with the known values, the average error of measurement is all less than 0.01mm, which meets the national regulations for overprint accuracy in the printing industry, and the detection time and detection efficiency can meet the real-time requirements.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261010 (2023) https://doi.org/10.1117/12.2671233
Aiming at the disadvantage of space manipulator under the working condition of large load and high precision positioning, such as the real kinematic parameters of space manipulator do not match the kinematic parameters of theoretical design, a method for on-orbit calibration of space six-axle manipulator is put forward. Based on this method, a kinematic calibration model is established, kinematic modeling and parameter identification analysis of the manipulator are completed, and accurate kinematic parameters can be obtained. Then the accurate end position and attitude are calculated, and calibration experiments are carried out. After the kinematic self-calibration experiment, After the kinematic self-calibration experiment, the pose accuracy of the end of the space manipulator is significantly improved at 60 test poses, which verifies the effectiveness of the proposed method. The experimental results show that the calibration method of space arm motion parameters proposed in this paper is correct, the physical meaning of the calibration model is clear, and the position and attitude positioning accuracy of the end of the arm is guaranteed.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261011 (2023) https://doi.org/10.1117/12.2671374
With wide-spread usage of style transfer, numerous methods for style transfer draw an increasing attention. Several methods to enhance the efficiency of style transformers have been made, one of them is StarGANv2, a method for multiple-style transfer, which can transform a batch of source pictures into other pictures with different styles. The main difference of StarGANv2 with other style transformers is that it uses style code to represent the styles to enable StarGANv2 to complete multiple-style transformation. The authors of StarGANv2 use CelebA-HQ and AFHQ dataset to train the model and test the model, and the results are pretty better than other style transformers. The goal of this paper is to exploit the effectiveness of StarGANv2 in the real-world scenes, such as over exposure or the angle facing the camera. The results validate the power of StarGANv2 where the model is robust enough to transfer the pictures into other styles. To achieve this, the authors of StarGANv2 use the photo clipped in videos which record real-world animals and form a new dataset. Then, the authors of StarGANv2 use the dataset to test the pre-trained model which is trained by AFHQ dataset and evaluate it according to FID metric. The authors of StarGANv2 draw a conclusion that StarGANv2 is robust in real world scenes. The meaning of this paper is that the authors get the real-world usage of StarGANv2 and have a test of StarGANv2’s robustness in real world photos and validate the potential of StarGANv2 in real-world applications.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261012 (2023) https://doi.org/10.1117/12.2671041
Accurate short-term traffic flow prediction is the basis and key to the intelligent transportation system. With the continuous development of machine learning algorithms and the latest swarm intelligence algorithms, a reasonable combination of the two will produce a good prediction effect. In this paper, BP neural network algorithm in the short-term traffic flow prediction problem accuracy is not high and easy to fall into the local minimum and so on. This paper established a BP based on adaptive BWO optimization short-term traffic flow prediction model, first of all, to carry on the data preprocessing the data set and divided into the training set and test set, and then the data for training, the best model to forecast practical optimization results, finally the model prediction results were compared with the rest of the 6 kinds of classical model. The experimental results show that the optimized BP model based on adaptive BWO can achieve a good traffic flow prediction effect in the short term, MAE is 7.357, MSE is 102.772, and R2 is 0.889, which are better than the other six models.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qingbo Ji, Jiangjiang Wu, Deqiang Kong, Lei Zhang, Changbo Hou
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261013 (2023) https://doi.org/10.1117/12.2671252
Glass tubes for vaccines have strict production requirements, and large-scale production is difficult to ensure their high quality. Minor damage or defects may seriously affect the quality of vaccines. In the actual task of vaccine glass tubes defect detection, there are many kinds of defects, and the defect forms of the same kind are also very different. Traditional manual detection is time-consuming, laborious and expensive. The traditional digital image processing algorithm is also poor in the field of defect detection. This paper proposes a DD-STDC net (STDC net for defect detection) based on semantic segmentation. Multi-scale input branches are introduced on the basis of the original STDC net, so that the model can learn the regional relationship between different scales through attention. For all kinds of irregular defects, deformable convolution is used in the backbone to capture the edge details of irregular objects more easily. The attention mechanism is embedded in the backbone network to strengthen the attention to the effective spatial location and the effective semantic information channel. On the test set, we achieve 66.3% MIOU on NVIDIA RTX 2060Super, which is 1.55% higher than the STDC net.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261014 (2023) https://doi.org/10.1117/12.2671258
In the expressway network, a large number of abnormal vehicle overtime behaviors occur every day. Currently, there is no efficient detection method. To solve this problem, this paper presents a vehicle full probability prediction optimization model, which can calculate the abnormal probability of other vehicles affected by the event after one vehicle is identified as abnormal vehicle. Further on, this paper establishes multiple events linear estimation model for two probability vehicle types. Finally, this paper presents a probability algorithm of abnormal overtime driving behavior to calculate the abnormal probability of various probability vehicle types.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261015 (2023) https://doi.org/10.1117/12.2671474
Liquid Composite Molding (LCM) processes are more cost-effective compared to autoclave-cured prepreg, but an independent preforming step is typically required to convert 2D fabric blanks into complex 3D shapes prior to molding. Numerical models are therefore important to predict the formation of defects during the design phase, in order to ensure the quality of final composite components. A macroscopic finite element model was employed to predict the forming behavior of multi-layered biaxial Non-Crimp Fabrics (NCF) during the press tool forming using a hemispherical punch. The forming behavior of the NCF was predicted by simulations considering the bending stiffness of the NCF, enabling fabric wrinkling to be simulated. Simulation results indicate a correlation between fabric wrinkling and the in-plane shear deformation of fabrics. The severity of wrinkles was also influenced by the layup sequence. Compared with the single orientation layups, more wrinkles were predicted for the layup comprising plies stacked at different orientations due to the dissimilar shear deformation between these plies.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261016 (2023) https://doi.org/10.1117/12.2671217
Users of the simulation software not only need to model the capability of each unit, but also need to create a decision maker for the units in the simulation software, to control ships, aircrafts and ground units to cooperates to achieve one goal. In this paper a new approach is constructed to create the decision maker. We use reinforcement learning based on global critic and local actor. The invention constructs an air isomorphic formation command method based on multiagent PPO algorithm. The evaluation network uses global information, so that the algorithm has the ability to evaluate global information and guide the agent to select actions that are beneficial to the global environment state. The input of the action network is local information, so that the agent can focus on local information.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261017 (2023) https://doi.org/10.1117/12.2671141
Water quality testing has important application value for agricultural ecological restoration. It is of great significance for the growth of farm crops and the protection of ecological environment through the detection and management of water quality in farm ponds. In order to realize the effective development of water quality detection of unmanned ships, solve the problems of immature monitoring system technology and poor real-time data of small, unmanned ships, a three-dimensional visualization information system for water quality detection unmanned ships was designed based on virtual reality technology. The test results showed that the system has good real-time performance and stability, which proves the feasibility of the system, and provides a new idea for the design and development of unmanned ship remote monitoring system.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261018 (2023) https://doi.org/10.1117/12.2671363
With the popularization of the Internet and the rapid development of artificial intelligence, the requirements for the accuracy of sentiment analysis on text data have been continuously improved, and aspect-level sentiment analysis has emerged as the times require. This paper analyzes and combines the research results of aspect-level sentiment analysis in recent years and proposes an aspect-level sentiment analysis model integrating GPT and multi-layer attention. (GPT and Multi-Layer Attention Network, GPT-MAN). The model first combines GPT and aspect coding to obtain a new word vector representation method, and then inputs the word vector with rich semantics into the network with a multi-layer attention mechanism, Sentiment Analysis Accuracy. In the experiment, the GPT-MAN model is analyzed from the selection of word embedding model and the optimization of hyperparameters using the datasets of Restaurant, Laptop and Twitter, and the effectiveness of this method is verified. And in comparison, with different models, the accuracy of the model in the three datasets increased by 2.57, 3.34, and 1.19 percentage points, respectively.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261019 (2023) https://doi.org/10.1117/12.2671081
The cell microinjection task in space requires the operator on the ground hold the handle of haptic device to control the remote dexterous manipulator in Space Cabin to needle into the cell for gene injection or nucleus extraction. To prevent the failure of punctures, the reliable force feedback plays a key role to adjust the position and velocity of the needle of manipulator in the process. In this paper, a feasible haptic rendering approach is presented to carry out the cell microinjection teleoperation. The cell is modeled as a sphere-tree adjacently connected with deformed springs with its cytomembrane and inner nucleus physical properties. A configuration-based constrained optimization method is performed to calculate the feedback force. We also propose a locking method to maintain the force feedback stable when the needle passes through cell boundaries with different physical properties. Finally, three sets of experiments are designed to validate the efficiency and stability of our method in cell microinjection.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101A (2023) https://doi.org/10.1117/12.2671445
With the rapid development of digital technology and the popularization of terminal equipment, the new media academic community is booming. The knowledge information in the academic community is mainly generated by the user 's posting and replying, so the quality of academic resources is closely related to the creator 's knowledge level and cultural background. However, the user level is mixed and difficult to evaluate, which makes the academic community user generated content have high redundancy and low quality, and seriously reduces the efficiency of academic users to acquire knowledge. Therefore, this paper uses Word2Vec word embedding model and professional domain dictionary to vectorize and automatically label user-generated content, and then trains the Bi-GRU neural network model to construct the quality evaluation method of user-generated content, which provides a basis for the quality identification and evaluation of user generated content in the new media academic community.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101B (2023) https://doi.org/10.1117/12.2671147
In this paper, three convolutional neural network models are used to achieve end-to-end recognition of blood cell images. The network model parameters are initialized by transfer learning from the pre-trained model on ImageNet, and then the blood cell images are input into the model, and the network model training is completed by back-propagation to continuously update the parameters. For small-scale datasets, the number of blood cell images is expanded using data increments to improve the generalization ability of the model. Experimental results on the BCCD dataset show that the best result MobileNetV2 achieves an accuracy and precision of 0.894 and 0.916, respectively.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101C (2023) https://doi.org/10.1117/12.2671434
Compared with the general Visual Question Answering (VQA), Medical VQA is more challenging. Medical images contain more complex information than general images. Aiming at this point, we propose the IIF module that can improve the model's ability to obtain visual feature. In addition, we design QAM to help the model analyze the question better. On the VQA-RAD dataset, the accuracy of our model improved to 66.4% on the opened-ended questions and 80.1% on the closed-ended questions, outperforming other relevant models. The results on the VQA-MED 2019 dataset also verify the effectiveness of our model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101D (2023) https://doi.org/10.1117/12.2671165
At present, Chinese named entity recognition techniques incorporating lexical knowledge have gained good development on graph neural networks. However, since there is still a deficiency in graph neural networks for longer distance semantic information and its pointing and localization information. In this task, we try to extract the missing pointing and locating information by employing relative position encoding techniques; it deals with long-distance dependencies by obtaining lexical information through dependency parsing. It is also able to deal more effectively with long-distribution relations and missing data in terms of direction and location through interactions with words in relative location coding. The experimental results on three NER datasets show that the proposed model is improved compared with other governor models.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101E (2023) https://doi.org/10.1117/12.2671421
Single image super-resolution is an approach to optimize the image stripe structure and improve the image quality. Recently, it developed rapidly based on convolution neural network, specially designed for this task, which becomes a hot topic of research and have shown remarkable result. Recently, many models have been developed based on Generative Adversarial Networks (GAN) and display enormous superiority compared with traditional deep learning methods. In GANs settings, adversarial loss pushes the generated image to natural image manifold with the help of a discriminator and at the same time trains discriminator to better discriminate the real image from those fake images generated by generator. In this course of confrontation, the generator is excellently trained and have achieved outstanding performance in the image super-resolution task. However, the traditional SRGAN image super-resolution reconstruction algorithm has slow training convergence speed. Moreover, excessive high-frequency texture sharpening leads to distortion of some details, which has a negative impact on the reconstructed image. In this work, curriculum learning algorithm is implemented to solve these problems and thus originally propose CL-SRGAN method, which is designed to help SRGAN achieve better performance on image resolution task. In the final experiment, CL-SRGAN has made an effective breakthrough in processing image reconstruction.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101F (2023) https://doi.org/10.1117/12.2671092
Humans can quickly and efficiently extract information from a complex natural scene. Rapid detection of animals is such an example, which is fast and accurate. We can see that animals have gender differences, and human beings also have gender differences, and they all appear in our real life. Therefore, we will use a two-alternative forced-choice paradigm (2AFC) to investigate the gender differences between the two targets. In our experiment, we balanced the various factors that could be taken into account and subjected the images to histogram equalization. We analyzed the reaction time of the subjects to stimuli of the target gender (male or female). We report two main findings. First, when the type of target (human or animal) was not considered, subjects had faster reaction times to male targets than to female targets. Second, gender differences were only significant for animals when the kind of object (human or animal) was considered.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101G (2023) https://doi.org/10.1117/12.2671087
Reliability Centered Maintenance (RCM) technology can improve the reliability of equipment maintenance, but it still has problems such as low analysis efficiency and poor description completeness. We propose to construct a RCM maintenance strategy model based on logical reasoning. This model uses Answer Set Programming (ASP), a non-monotone logic programming language, to realize the theoretical modeling of RCM in the form of logical rules. We use preference optimization to improve the RCM analysis method and integrate CWA (Closed World Assumption) and NAF (Negation As Failure) into the ASP program. Practicality and generality are the main core objectives of this model. Finally, the turbine engine failure of aircraft is taken as the main research example. The effectiveness and efficiency of the model are verified by a comparison of model conclusion consistency. Experimental results show that compared with other RCM systems, this model has good efficiency, reliability, and completeness.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Hongwei Liu, Jinpeng Wang, Masud Talukdar, Da Yuan
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101H (2023) https://doi.org/10.1117/12.2671696
Ground Penetrating Radar (GPR) is usually used to detect unknown underground structure information, but it is very difficult to extract the underground target structure information from original GPR signal. This paper aims to solve the inversion problem by deep learning method. The two-dimensional ground penetrating radar B-scan signal is converted into intuitive underground structure information by neural network. In this paper, the DeepLab network proposed by Google is improved to solve the problem of permittivity inversion of GPR signal images. We verify the network using simulated data, which is generated by Finite Difference Time Domain (FDTD) algorithm. Finally, we quantitatively evaluate the performance of our network by comparing it with some existing deep learning inversion networks.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101I (2023) https://doi.org/10.1117/12.2671230
The painted murals in Mogao Grottoes and Longmen Grottoes are symbols of China history and culture. However, most of the murals with complex texture and structure have suffered from different degrees of disease erosion after thousands of years. It is necessary to restore the damaged parts of the murals and to accurately restore their contents. In recent years, the use of new virtual technologies such as digital images to repair the damage can largely avoid secondary damage to the murals caused by manual restoration methods. Therefore, this paper takes the restoration of the most typical shedding diseases to the Mogao Caves murals in Dunhuang as an example. Furthermore, the research object of this paper is the shedding diseases including contour lines. For the traditional virtual methods of repairing shedding diseases, the structure and texture are usually restored at the same time, and these methods have little effect on the accurate removal of shedding disease through the contour line. It can be seen that shedding disease through the contour line is more difficult to repair, and more appropriate inpainting methods need to be explored. Considering the particularity of the shedding disease that passes through the contour line, this paper proposes a mural image inpainting algorithm based on structure priority to repair the shedding diseases. First, the structure repair problem is further converted into a optimization problem, and then the global optimization capability of the genetic algorithm is used to realize the connection of the structure information of the damaged area. Then, the texture is filled by subarea optimization to obtain an ideal repair effect, which can reasonably and effectively solve the problem of shedding disease repair through the contour line. Subjective and objective evaluation of experimental results is also better than other comparative methods.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101J (2023) https://doi.org/10.1117/12.2671157
Low-light images suffer from low visibility, much noise, uneven illumination distribution, etc. Many existing methods have problems such as over enhancement or insufficient detail enhancement when dealing with low-light images with uneven illumination distribution. To remedy the above shortcomings, we propose a Retinex-based self-supervised low-light image enhancement model (Retinex-SIE), which is mainly composed of three parts: Retinex-based self-supervised image decomposition network (Retinex-DNet), nonlinear conditional illumination enhancement function (NCIEF), and Image Reconstruction (IR). First, a uniform illumination image of the same scene with the low-light image is generated by homomorphic filtering transformation, and the low-light image and the uniform illumination image are input into Retinex-DNet for decomposition to obtain reflectivity, noise and illumination. Then, NCIEF is used to enhance the illumination after decomposition. Finally, the final enhanced image is obtained by multiplying the decomposed reflectance and the enhanced illumination. Experiments on severa challenging low-light image datasets show that Retinex-SIE proposed in this paper can better handle low-light images with uneven illumination distribution and avoid problems such as excessive enhancement or insufficient detail enhancement.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101K (2023) https://doi.org/10.1117/12.2671102
This paper uses the knowledge mapping software of CiteSpace to process the index data of anti-monopoly literatures that retrieval from the core collection of Web of Science, and the results show that seven references have higher co-citation frequency and centrality; the research hotspots in this area include antitrust law enforcement, Sherman Law, competition policy; the research frontiers include Internet search, antitrust arbitration, public enforcement, auction design and etc.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101L (2023) https://doi.org/10.1117/12.2671339
With the development and progress of medical imaging technology, the resolution of medical images has been increasing, and a variety of high-definition imaging modalities such as CT, PET-CT and MRI have emerged. U-Net network has the advantages of simple network topology and small training set data requirement; thus, the field of medical image segmentation uses it extensively. However, U-Net also has some problems, such as edge loss of segmentation results, long training time and single application scenario. For the medical image segmentation problem, this paper proposes a method that combines channel attention and spatial attention and uses an improved join strategy to join the network structure. To address the problem of insufficient data volume of medical images, this paper performs a data augmentation operation on the dataset with elastic deformation. In addition, we use a local-global training strategy to further improve the performance of training on medical images. When compared to the original U-Net, the Dice coefficient and IOU metrics are significantly better when utilizing the method suggested in this work. After extensive experiments, the strategy proposed in this study can achieve good outcomes when facing medical image segmentation problems and has great potential.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qingbo Ji, Qingquan Liu, Pengfei Zhang, Tingshuo Yin, Changbo Hou
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101M (2023) https://doi.org/10.1117/12.2671277
The number of chronic kidney disease in China has shown a rapid upward trend. The morphological analysis of urinary erythrocyte is the key to diagnose various types of chronic kidney disease. Therefore, this paper studies the classification method of urinary erythrocyte based on supervised contrastive learning. Aiming at the low spatial resolution of urine erythrocyte and the difficulty in feature extraction, this paper proposes a Small Resolution Residual Network (SRRN) structure model based on ResNet-50 model. A multi-contrastive loss function is proposed for the singularity of feature similarity measurement in supervised contrastive learning. Based on supervised contrastive loss function, a feature similarity measurement method based on Euclidean distance is added. Aiming at the low accuracy and recall rate of some categories in urine red blood cell classification, this paper introduces a weight balance mechanism in cross entropy loss and sets higher loss weights for more difficult categories. The accuracy of this method and ResNet-50 network model on the urinary erythrocyte dataset is 92.26% and 90.7% respectively, which shows the effectiveness of this method on urinary red blood cell recognition.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101N (2023) https://doi.org/10.1117/12.2671319
In order to solve the problem that the traditional feature matching algorithm has less premise number of feature points and poor matching ability under outdoor complex lighting conditions, an image matching algorithm based on color invariants in outdoor environment is proposed. Firstly, a feature matching algorithm with color invariants and Tanimoto similarity is designed based on Kubelka Munk theory. By introducing color invariants to distinguish the available feature areas in outdoor scenes, AKAZE (Accelerated KAZE) algorithm and SIFT (Scale invariant Feature Transform) algorithm are combined to generate more comprehensive feature descriptors; Then, Tanimoto similarity test is used to screen feature point pairs and random sample consensus algorithm is used to remove external points. According to the experimental results, under the same conditions, the improved algorithm obtains more effective feature points at the edge of the image and in the smooth area of the image. The average accuracy of the algorithm in outdoor environments reaches 90%, and the number of feature matching is 43% higher than that without color invariants.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101O (2023) https://doi.org/10.1117/12.2671190
A software modeling method for hot working domain process systems based on domain metamodeling is proposed to address the problems in the development of Hot working domain process systems. Combining metamodeling, domain modeling and MDA to realize model-based Hot working process system development, the abstract syntax, concrete syntax and semantics of the domain modeling language are designed using MetaEdit+, a metamodeling tool. And based on this, a heat treatment process model was established. The method effectively improves the efficiency and model reusability of process system development software in the hot working domain.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101P (2023) https://doi.org/10.1117/12.2671204
In the modeling work of Knowledge Graph Completion (KGC), we propose a KGC model considering both entity embeddings and relational paths for small sample data. The relational path information between entities can identify the relative positions of two entities in the knowledge graph, and if the distance is too far from there is no need for link prediction with high probability, so limiting the number of hopes of relational path can reduce the cost of link prediction. There are many link prediction models that only consider relations, but such models are not suitable for small sample datasets because there are too few types of relations in small datasets, and considering only relations is not a good way to characterize entities, so we added entity embeddings to consider relational paths, aggregated entity neighborhoods and relational neighborhoods around entities to target entities, and finally combined entity embeddings with relational paths to perform link prediction tasks. We have tested on three small sample datasets and achieved remarkable results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101Q (2023) https://doi.org/10.1117/12.2671262
Low-light image enhancement is to restore the image acquired under insufficient light conditions to the normal exposure image. The low-light image enhancement method based on Retinex theory is a common method. The image is decomposed into the light component and the reflection component, and the corresponding enhancement is done respectively and then fused to achieve the purpose of image enhancement. However, most of the commonly used decomposition and enhancement networks in this field are designed by stacking convolution or up/down sampling, which lacks the guidance of relevant semantic information, resulting in the loss of many details in the decomposed and enhanced images. In order to alleviate the above problems, we propose a low-light image enhancement model based on Retinex theory and residual attention. Under the guidance of semantic information provided by channel domain and spatial domain, it can obtain smoother and less noisy images in the decomposition stage. In the image enhancement stage, the image texture and color can be restored with high quality. Moreover, we design loss functions that are more suitable for decomposition and enhancement tasks to constrain the learning tasks of different networks. In addition, we designed a residual block fused with dual attention unit, which can stably extract richer image features and suppress the generation of noise. Finally, we compare our model with the mainstream methods in recent years on public datasets. Extensive experimental results show that our model is superior to the mainstream methods, showing excellent performance and potential.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101R (2023) https://doi.org/10.1117/12.2671468
Nowadays, target recognition, driverless, medical impact diagnosis, and other applications based on image recognition in life, scientific research, and work, rely mainly on a variety of large models with excellent performance, from the Convolutional Neural Network (CNN) at the beginning to the various variants of the classical model proposed now. In this paper, we will take the example of identifying catamount and canid datasets, comparing the efficiency and accuracy of CNN, Vision Transformer (ViT), and Swin Transformer laterally. We plan to run 25 epochs for each model and record the accuracy and time consumption separately. After the experiments we find that from the comparison of the epoch numbers and the real-time consumption, the CNN takes the least total time, followed by Swin Transformer. Also, ViT takes the least time to reach convergence, while Swin Transformer takes the most time. In terms of training accuracy, ViT has the highest training accuracy, followed by Swin Transformer, and CNN has the lowest training accuracy; the validation accuracy is similar to the training accuracy. ViT has the highest accuracy, but takes the longest time; conversely, CNN takes the shortest time and has the lowest accuracy. Swin Transformer, which seems a combination of CNN and ViT, is most complex but with ideal performance. In the future, ViT is indeed a promising model that deserves further research and exploration to contribute to the computer vision field.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101S (2023) https://doi.org/10.1117/12.2671731
The fisheye camera is widely used in computer vision because of its large field of view. However, in optical theory, the large field of view is at the cost of distortion. It is impossible to obtain effective information directly from the fisheye image, so the original image needs to be distorted first to become a linear image without distortion. Aiming at the existence of traditional longitude correction in the image center, edge correction effect is different and edge distortion, introducing the repositioning center algorithm and stretch factor. Firstly, the effective area of fisheye image is obtained by row-by-row column scanning method. And then uses the repositioning center algorithm to obtain the new center and radius, according to the distortion principle to calculate the distortion principle. The mapping relationship between the target image and the original image is obtained, and finally modified by the bilinear interpolation method. Compared with the traditional longitude correction, the proposed algorithm can correct the fisheye image accurately and effectively and improve the quality of correction.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101T (2023) https://doi.org/10.1117/12.2671788
Global pandemic due to the spread of COVID-19 has post challenges in a new dimension on facial recognition, where people start to wear masks. Under such condition, the authors consider utilizing machine learning in image inpainting to tackle the problem, by complete the possible face that is originally covered in mask. In particular, autoencoder has great potential on retaining important, general features of the image as well as the generative power of the Generative Adversarial Network (GAN). The authors implement a combination of the two models, context encoders and explain how it combines the power of the two models and train the model with 50,000 images of influencers faces and yields a solid result that still contains space for improvements. Furthermore, the authors discuss some shortcomings with the model, their possible improvements, as well as some area of study for future investigation for applicative perspective, as well as directions to further enhance and refine the model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Yan Zhou, Haohai Wu, Xiangyu Liu, Fanzhi Zeng, Yuexia Zhou
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101U (2023) https://doi.org/10.1117/12.2671328
With face recognition playing a crucial role in biometric identification technology, Face Anti-Spoofing (FAS) has powerful effects on finding out whether a presented face is live or spoof. As the most common attacks such as photo attacks, print attacks, and video replay attacks can be effectively resolved, high-resolution attacks are easy to occur but still challenging for effective face spoofing because of the rich local facial details. In this paper, a Diagonal-Fusion Transformer network (DFT) which adds self-attention from the vision transformer is proposed. It is designed to learn the facial context information and relation between the local features of the face, and thus enhance the discriminative features of the real face and the fake face to improve the classification efficiency. Furthermore, a Spoofing Region Detection network (SRD) parallel with the DFT network is proposed for fine- grained spoof detection through the enlargement of local facial details. Through comprehensive experiments, the model achieves state-of-the-art results on public benchmark datasets such as OULU and CelebA-Spoof.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101V (2023) https://doi.org/10.1117/12.2671148
To improve the real-time performance and robustness of traditional feature matching algorithms, an improved image feature matching algorithm E-OrbF based on ORB and FREAK is proposed. In E-OrbF, the original FAST feature points in ORB algorithm are distributed unevenly and redundant. The strategy of subregion and local threshold is adopted to improve the uniform distribution and stability of feature points. Then simplify the sampling mode of FREAK algorithm and design a new feature descriptor. While improving the matching speed, the sampling point pairs are further filtered to improve the matching accuracy. Finally, combine RANSAC matching algorithm to eliminate mismatches and reduce the rate of mismatches. The experimental results show that the algorithm has good real-time performance, while under the conditions of perspective transformation, rotation scale, complex illumination and blur. Both of them can well complete feature detection and feature matching and improve the robustness of existing methods. The algorithm can be applied to the fusion of virtual and real scenes on mobile terminals, and the average visual frame rate reaches 30 FPS, meeting the real-time requirements.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101W (2023) https://doi.org/10.1117/12.2671439
As we know, recently deep learning networks have gained currency for some time under the background of the rise of big model, and it has been widely used for various areas including microbiology images recognition. Nowadays deep learning network models are also divided into many different types, and many new models are proposed every year to achieve better performance. After introducing their specific principles and composition structure, this paper compares three common different deep learning networks named Deep Neural Network notation (DNN), Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) and introduces the recently emerging model named Residual network (Resnet), starting from a specific situation which calls for garbage classification. Moreover, during this experiment, it also gives the method V pipe a try which differs a lot from the former method and receives good results worth celebrating.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Chunli Wang, Botao Zeng, Jindie Gao, Ge Peng, Wei Yang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101X (2023) https://doi.org/10.1117/12.2671074
In recent years, the traffic image semantic segmentation plays a crucial role in automatic driving. The result of semantic segmentation will directly affect the car's understanding of the external scene. Thus, a semantic segmentation algorithm based on UNET network model is proposed for getting better results in traffic images segmentation. To prove the effectiveness of the proposed algorithm, highway driving dataset is used on the experiments. The experimental results show that the proposed network can achieve high precision image semantic segmentation in complex road scenes, and the segmentation accuracy is greatly improved compared with other network models.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101Y (2023) https://doi.org/10.1117/12.2671332
Mask R-CNN is a relatively mature method for instance segmentation at this stage, aiming at the problems of segmentation boundary accuracy and poor robustness to blurry pictures in the Mask R-CNN algorithm, an improved Mask R-CNN instance segmentation method is proposed. Use the SegNeXt network structure to optimize the mask branch for further segmentation of candidate regions, and then use new anchor size and IOU standards so that the candidate box can cover all instance regions. Finally, a method is used to add partially transformed data from a transformation network for training. Compared with the original algorithm, the total mAP value is increased by 2.2%, and the accuracy and robustness of the segmentation boundary are improved.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126101Z (2023) https://doi.org/10.1117/12.2671228
In order to solve the problem of low accuracy of road network information extraction in high-resolution remote sensing images due to complex ground object environment, this paper proposes an improved deep learning semantic segmentation model CP-Unet. In this model, the CBAM full-connection layer module is used to enhance the feature fusion of the model. At the same time, the subpixel convolution up sampling module is introduced to reduce the loss of definition caused by the amplification of the dimension of the feature map in the up sampled convolution. Finally, the model is more suitable for road network extraction in high-resolution remote sensing images. In order to verify the reliability of CP-Unet model, an area of Xinjiang Road network was taken as the object of the experiment. The overall extraction accuracy index IoU score of the model in this paper is 81.73%, which is 6.66% higher than that of U-Net. It can better overcome the complex environmental interference and extract the road network in a more complete way. It provides method reference for road network information checking and updating.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261020 (2023) https://doi.org/10.1117/12.2671267
Deep learning segmentation networks can effectively solve the crop disease segmentation problem, but maize diseases usually vary greatly in size, leading to their unsatisfactory segmentation accuracy. To address this problem, this paper proposes an improved Deeplabv3+ segmentation network model. First, in the encoding stage, feature extraction using Resnet101 with atrous convolution, and the Jump-Connected Atrous Spatial Pyramid Pooling (JCASPP) module is designed to obtain multi-scale semantic information. Second, in the decoding stage, the JCASPP output is fused with the shallow features of the backbone network to obtain richer spatial information by using multilayer and small multiplicative up sampling. The comparison experimental results with the traditional DeepLabv3+ model and its two improved models show that the segmentation accuracy of this model is higher, and the average cross-merge ratio reaches 77.3%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261021 (2023) https://doi.org/10.1117/12.2671208
Paper, glass, fruit, scrap metal equipment, and other materials have become domestic garbage as people's living standards have dramatically improved. Plastic trash, including bags, and disposable lunch boxes, has become a global environmental threat. Carrying out waste separation and encouraging waste reduction at the source is a meaningful way to realize the quantification, harmlessness, and resourcefulness of domestic waste. Pictures contain lots of information data and many methods for successfully extracting and examining that information data. Consequently, one of the primary research projects in the field of pictures is the picture classification problem. Traditional image classification techniques can no longer process and extract information from vast amounts of image data quickly. Throughout the experiments, the accuracy of ResNet-50 is lower than ViT, most likely due to its less extensive set of parameters. And the more the parameters of the model, the higher the precision of training. Additionally, the experiment examined the performance of the ResNet-50 model with and without pretraining, and the results showed that the performance of the model without pretraining was noticeably inferior to the model with pretraining. It is better to have a model which has numerous parameters.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261022 (2023) https://doi.org/10.1117/12.2671240
We propose a face image restoration method that combines feature fusion and attention mechanisms for the current image restoration field that generates blurred images, artifacts, inconsistent texture and structure fusion. The model divides image restoration into two stages. First, the edge information repaired by the edge generation adversarial network is used as the prior knowledge of the image, and then the generated prior knowledge and the broken image are put into the image repair network to generate the complete image. We introduce a texture-structure feature fusion method in the generator structure to solve the texture and structure fusion inconsistency problem and use a dense residual layer-hopping connection to mitigate the gradient disappearance problem while speeding up the model convergence and introduce a spatial and channel attention mechanism to generate correct semantic connections to enhance the model performance and suppress image blurring. We apply the algorithm to the CelebA-HQ face dataset, and compared with the current mainstream restoration algorithms, quantitative analysis shows that the method in this paper outperforms in three metrics, PSNR, SSIM, and L1.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261023 (2023) https://doi.org/10.1117/12.2671070
This paper proposes an electrical capacitance tomography algorithm based on an elastic network. To obtain feasible solutions, the L1 and L2 norms are used as the regular terms of the objective function, so that the solution has both the feature selection characteristics of the L1 norm and the image smoothing characteristics of the L2 norm. And utilize the normalized Laplacian as the weight of the elastic network, perform edge detection, and identify the dominance of L1 and L2. This algorithm makes the imaging region smooth, preserves the edge details of the image, and increases the accuracy of the image.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261024 (2023) https://doi.org/10.1117/12.2671523
The digitalization of painting works is of great significance to the effective use of painting resources. Traditional image classification methods do not consider the subjective characteristics of painting works, and most of the features need to be manually extracted. There are problems such as loss of detail features. In this paper, a painting image classification method based on convolution neural network is proposed, and the influence of the size of convolution kernel, the structure width of convolution neural network, and the number of training samples on the classification results is analyzed to optimize the network structure and parameters. The experimental results show that the method is effective for the classification of painting images, and the classification results of different painting image data sets are also good.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261025 (2023) https://doi.org/10.1117/12.2671043
In the big data environment, the key is the precise recommendation of learning resources to learners. The core is the in-deep mining of learners’ personalized demands. This study solves this problem by constructing learner personas. Primarily, collect web learning data of learners to cluster them. Then analyze the characteristics of learners to predict their learning intentions and knowledge blind spots. Based on it, generate a clear personalized learning path subsequently. Precise positioning, quickly finding out the learner's ability and quality shortcomings. And completing the accurate recommendation to learners. It will help learners establish a reasonable learning path, and provide more accurate service support. This study will provide a theoretical basis for carrying out big data precision services and meeting the personalized learning needs of learners.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261026 (2023) https://doi.org/10.1117/12.2671199
In order to enhance the ability to diagnose and distinguish COVID-19 from ordinary pneumonia, and to assist medical staff in chest x-ray detection of pneumonia patients, this paper proposes a COVID-19 x-ray image detection algorithm based on deep learning network. First of all, a model of deep learning network is set up based on VGG - 16, and then, the network structure and parameter optimization is adjusted, which makes the network model can be applied to COVID-19 x-ray imaging detection task. In the end, through adjusting the image size of the original data set, the input data meets the requirements of the deep learning network. Experimental results show that the proposed algorithm can effectively learn the characteristics of the COVID-19 x-ray image data set and accurately detect three states of COVID-19, common viral pneumonia and non-pneumonia, with a very high detection accuracy of 95.8%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261027 (2023) https://doi.org/10.1117/12.2671318
Neurological diseases, including Alzheimer's disease and brain tumors, are the leading causes of death and disability worldwide. However, it is difficult for scientists to quantify the response of these deadly diseases to treatment. Existing neuron-based solutions have limited accuracy. Neuroblastoma cell lines have unique, irregular and concave morphology, which makes them show low precision scores in different cancer cell types. Based on this, this study proposes a new cell semantic segmentation network model. The model first enhances the original cell map, and then introduces the residual module and attention mechanism based on the classical U-Net network structure, which alleviates the problem of network degradation and improves the efficiency and effect of network training. The experimental results on the neuroblastoma cell line data set provided by Sartorius show that the segmentation accuracy of the proposed model is about fifteen percentage points higher than that of the classical U-Net model and one percentage point higher than that of the U-Net++ model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261028 (2023) https://doi.org/10.1117/12.2671142
Lithium-ion battery has become an indispensable energy storage component in our life because of its environmental protection and high energy characteristics. The battery SOH is the decisive factor to ensure its stability. For the sake to improve the accuracy of EV battery SOH prediction. Firstly, data structuring, PCA dimension reduction and data standardization were used to transform downloaded data into data that could be trained with high accuracy model. After that, the characteristic factors related to battery capacity were extracted from the battery charging data and correlation analysis was carried out. According to the method of Pearson coefficient, the features with strong correlation were left and then imported into the sample data. The factor parameters of SVR and other models were optimized by grid search algorithm, and the final prediction model was established. Lithium-ion battery has become an indispensable energy storage component in our life because of its environmental protection and high energy characteristics. The battery SOH is the decisive factor to ensure its stability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261029 (2023) https://doi.org/10.1117/12.2671210
A chain pitch measurement method of scraper conveyor based on speckle structured light is proposed to improve security and automation. Two speckle structured light cameras are used to collect point clouds on the chain surface from different angles, and the point clouds are transformed to a common reference coordinate system by rotating and translation matrix with calibration. The chain point cloud is preprocessed by plane model segmentation and radius filtering, the main direction of the point cloud is calculated by point cloud principal component analysis, the key points of measurement are detected by neighbors within radius search of the point cloud, and finally, the chain pitch is solved by Euclidean distance. The actual measurement error of the measurement method proposed in this paper is less than 2%, which can meet the needs of coal mining.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102A (2023) https://doi.org/10.1117/12.2671195
Expressway project is usually built in extremely complex natural and cultural environment. The whole process of project implementation management is a continuous and dynamic management practice process, which will be affected by internal and external uncertainties, and may directly affect the benefit and even the survival and development of enterprises. Therefore, this paper studies and analyzes the risk of investment in the highway project and several factors that may affect it. This paper selects the actual situation of 112 expressways in China and analyzes them through 30 different risk indexes. Through constructing multiple linear regression model, the factors that may affect the investment risk of expressway project are analyzed. Finally, there are 20 risk indicators to influence the investment risk of expressway project, and this paper constructs the weight model of expressway investment risk evaluation hierarchy and tries to verify it.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102B (2023) https://doi.org/10.1117/12.2671431
In the Chinese medical question and answer task, question intention detection is a very important part. At present, the common intention detection methods mainly use the manually designed matching rules to find the problem features to detect the intention of the problem, but the use of a large amount of labor usually brings about problems such as high cost and poor versatility. A novel method of intention detection is proposed in this paper. First, the collected questions with different intention categories are used to construct intention feature words. Then, based on the BERT pre-training language model, a two-classification model of phrase similarity is constructed. By comparing the combination results of problem word segmentation and the similarity of intention feature words, the multi-classification problem of problem intention detection is transformed into a two-classification problem between multiple phrases. Then we can get the inclination of the question for each intention category, that is the intention category of the question. The experiment shows that the method based on the two-classification model of phrase similarity has better effect than the previous methods, and the F1 value in the test set reaches 90.1.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102C (2023) https://doi.org/10.1117/12.2671161
As size of pin arrays increases significantly, Printed Circuit Board routing becomes an arduous and challenging problem. Traditional manual routing might cost a lot of time. Simultaneous escape routing is a routing problem based on multiple pin arrays, and it’s an important stage of PCB routing. In this paper, we propose a novel algorithm to solve the SER problem based on Multi-Commodity Flow. Experimental results show that our method can obtain shorter wire length and use less boundary resources than other methods.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102D (2023) https://doi.org/10.1117/12.2671188
With the wide application of deep learning, the abstractive text summary has become an important research topic in natural language processing. The abstractive text summary has high flexibility and can generate words that have not appeared in the text. However, the generated summary model will have factual errors, which significantly affect the usability of the summary. Therefore, this paper proposes a text summary model based on fact relationships and keyword fusion. We extract the fact relation triplet in the input text and automatically extract the keywords in the text to assist in the generation of the abstract. The fusion of fact relations and keywords can effectively alleviate the problem of factual errors in the abstract. Many experiments show that compared with other baseline models, our model (FRKFS) improves the performance of summaries generated on the data sets CNN/Daily Mail and XSum and alleviates the problem of factual errors.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Chengwei Kang, Peicheng Cong, Yongbo Sun, Shengqi Wang, Xi Liu, Longjie Duan, Kuan Wu, Peng Cao, Dong Qin, et al.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102E (2023) https://doi.org/10.1117/12.2671108
With the rapid development of science and technology and the advent of the information age, the number of components used in electronic devices has increased sharply, making its internal circuit structure increasingly complex. Printed Circuit Boards (PCBs), as part of electronic devices, are becoming smaller and more integrated, resulting in a much greater increase in the probability of failure and the difficulty of detection. Therefore, to reduce the difficulty and cost of PCB fault diagnosis, it is very necessary to explore and study new PCB diagnosis methods. This paper first reconstructs the PCB dataset by ESRGAN, and then the CenterNet based on the center point is introduced and improved. The ResNeSt based on the segmentation attention mechanism is integrated with CenterNet to realize the PCBs fault diagnosis method based on the tiny object detection method. Experiments have proved that the method can achieve 99.42% mAP.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102F (2023) https://doi.org/10.1117/12.2671170
In our present work, the sample entropy (SampEn) was used to analyze the changes of ship driver driving fatigue state when they drive for a long time from the perspective of human electroencephalogram (EEG). Combining with the eye movement, the law of fatigue change caused by long-time driving of the ship is analyzed. Thus, it can conclude that the EEG characteristics, as well as the eye movement, can effectively detect driver’s fatigue when they are driving.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Xuejiao Li, Zhiwei Cheng, Wei Wang, Yongsheng Deng
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102G (2023) https://doi.org/10.1117/12.2671219
Water property forecasting can provide decision support for the protection and management of water resources. A big data analysis model, Multi-scale Extreme Learning (MEL), is reported in this work to address water property forecasting. Based on the divide-and-conquer philosophy, ensemble empirical mode decomposition is first adopted to decompose the Total Phosphorus (TP) that is a representation of water property into multi-scale features. The extreme learning machine is then employed to establish regression models in different scales. The outputs of multi-scale regression models are finally summarized into the ensemble forecasting result. A time series of historical weekly TP is introduced to validate the proposed MEL. Experimental results reveal that the proposed model based on the multiple scales representation capacity and the non-linear mapping, therefore, has the best excellent performance in water property forecasting.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102H (2023) https://doi.org/10.1117/12.2671326
In order to resolve the high resource consumption, labor shortage and low information integration problems in the existing extensive construction mode of railway bridges, in combination with railway bridge engineering characteristics, after a general research on intelligent construction necessity and a variety of cutting-edge information technologies, this paper presents a railway bridge information construction management platform which is based on "cloud + end" system, combining with BIM, GIS, cloud computing, big data, Internet of Things, artificial intelligence and other technologies. The main technical route is as follows: in terms of functional architecture, the platform includes eight modules related to management process and digital construction, ensuring clients two interaction modes: web terminal and mobile terminal. In terms of logical architecture, the platform takes spring cloud as the technical framework, using spring boot to develop distributed systems with microservices architecture, and decomposing various business functions into multiple discrete services; In terms of physical architecture, the platform mainly includes front-end server, database server, application server, BIM model management server and firewall, all of which are deployed in Alibaba cloud. The platform is applied to the Changqing Yellow River Bridge Project of Zhengji Railway. Results show that this platform can significantly improve the production efficiency, management efficiency and decision-making ability on site, realize intelligent site with digital twin characteristics, ensure construction safety and save project investment.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102I (2023) https://doi.org/10.1117/12.2671040
With the rapid development of information technology, artificial intelligence technology and the financial industry began to deeply integrate up. Algorithmic trading, credit card fraud detection and a series of other new technologies being applied to the financial industry all require a large amount of data support. However, due to the increasing amount of online financial data, it is difficult for the majority of investors and financial industry practitioners to obtain the required information in a timely manner. Entity recognition technology, as the basis of natural language processing, can quickly extract effective information from the massive financial texts and can provide effective help for investors and financial industry practitioners. In this paper, we propose a neural network model based on Bert-BiLSTM-CRF, which is applied to recognize financial entities. Through experimental analysis, the model achieves more than 95% of all indicators. Compared with the conventional model, the model has superior performance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102J (2023) https://doi.org/10.1117/12.2672298
Under the background of the rapid development of artificial intelligence, intelligent natural language processing technology has made rapid development. Different from the question answering system in other fields, the establishment of electric power Intelligent Customer Service System (ICSS) makes the construction of intelligent question answering system more challenging. This paper discusses and analyzes the design and implementation of intelligent question and answer system, the demand analysis of power ICSS and the construction process of address and place database. Through the function, performance and security test of power intelligent customer service, the test results show that the version of the release system runs stably and has been actually applied in the management of power customer service. After the launch of the power CSS, customers generally reflect that the effect is very good. They agree that the promotion of the power CSS in the power industry is an extremely important basic project to realize the hierarchical requirements of standardized management and operation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qingbo Ji, Ziqi Li, Kuicheng Chen, Qingfeng Ma, Lei Zhang, Changbo Hou
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102K (2023) https://doi.org/10.1117/12.2671257
Object tracking is a very challenging branch in the field of computer vision and plays an extremely important role in different fields. However, most algorithms still have no good robustness when the target is out of view, fast motion and low resolution. Therefore, this paper proposes a method based on Siamese architecture and optical flow to address these situations. The tracking framework is mainly divided into tracking network and target localization network based on optical flow. The tracking network uses SiamBAN and the target localization network uses LiteFlowNet3 to estimate the optical flow information of target. The proposed method in this paper expands the search area of SiamBAN, and uses the target localization network to reposition the target when tracking fails. The target localization network can be easily embedded into the existing Siamese network to provide more accurate location of target during the tracking process. Experiments results on VOT2016, VOT2018 show that this method has good robustness against extreme conditions.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102L (2023) https://doi.org/10.1117/12.2671042
Nowadays, safe driving has become a very serious problem all over the country and even the world. The consequences of most traffic accidents are human factors, such as tired driving, driving, making phone calls, talking back, etc. In order to reduce traffic accidents caused by human factors, based on the original convolutional neural networks (CNN), a driver behavior detection method based on convolutional neural networks and data enhancement is proposed in this paper. This method is based on the original convolutional neural network, optimizes the original convolutional neural network, and designs a model suitable for the task of this paper. At the same time, this paper introduces the data enhancement transformation method to expand the original data to improve the over fitting problem caused by the lack of data. Experiments show that the accuracy of this model is about 82.01% higher than that of the original model, and the results of using data enhancement are 4.12% higher than that of using original data.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102M (2023) https://doi.org/10.1117/12.2671281
Siamese network is successfully applied in object tracking. Most of the existing Siamese tracking methods extract template features in the first frame, which will cause the tracker to ignore the appearance change of the target in the subsequent video. In this paper, we propose a tracker based on foreground adaptive bounding box and motion state redetection. The tracker infers the reliability of tracking by the motion pattern of the bounding box. When an anomaly is detected, the tracker will redetect using the continuously updated template. Furthermore, our tracker employs an adaptive bounding box to avoid the effects of inaccurate rotation of the bounding box. The results on the VOT2018 dataset show that our tracker achieves stronger robustness and higher accuracy, providing superior performance compared to the current state-of-the-art trackers.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102N (2023) https://doi.org/10.1117/12.2671575
Micro-expressions are rapid, difficult to observe with the naked eye, facial expressions that can reflect real human inner emotions. Micro-expression recognition is still a great challenge due to the characteristics of very short duration and subtle changes (small amplitude of muscle contraction or diastole). Based on this, this paper proposes a 3D convolutional micro-expression recognition method based on attention mechanism, which is a dual-stream structure that can effectively utilize the features of image sequence and optical flow sequence. More effective micro-expression features are extracted using Attention layer, Co-Attention layer to better solve the micro-expression recognition task. Adequate experiments are conducted on the dataset to verify that the model has better recognition results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102O (2023) https://doi.org/10.1117/12.2671435
Dynamic gesture recognition is a very important interaction method in human-computer interaction. For the current research, multi-modal data and three-dimensional convolutional neural network are often used for training. Although the recognition accuracy is high and robustness is good, the amount of parameters is large and high computational cost. To solve the problem, a dynamic gesture recognition method based on Temporal Shift Module (TSM) is proposed on the basis of two-dimensional convolutional neural network. This method uses PyConvResNet-50 as the backbone network, adds the TSM module for information exchange in the time dimension, embeds the Motion Excitation module (ME) into the TSM to enhance short-term temporal modeling, and finally uses 2D-FCN for spatiotemporal feature fusion classification. The experimental results show that the recognition accuracy of the model on the large-scale gesture dataset Jester is 96.49%, which is comparable to that of the three-dimensional convolutional neural network, but the calculation amount is reduced by 63% as well. This method is suitable for the field of gesture recognition that requires high real-time performance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102P (2023) https://doi.org/10.1117/12.2671349
For embedded modern equipment, the current gait recognition algorithm model is difficult to deploy on it due to a large amount of gait frame image data, slow network processing speed, complex structure and low computational efficiency. In this paper, a lightweight convolutional network model integrating the attention mechanism is proposed. The algorithm first performs morphological processing on the image, extracts the gait contour image, and calculates the gait energy image; integrates the attention mechanism with MobileNetV1. The feature information of the image is effectively extracted, and the parameters of the network are reduced. A number of body method validation experiments are conducted in the CAISIA-B gait database of the Chinese Academy of Sciences, and the experimental results are significantly improved with other deep learning models.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qizheng Huo, Shaonan Li, Chengyang Li, Yongqiang Xie, Zhongbo Li
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102Q (2023) https://doi.org/10.1117/12.2671433
In the current cloud-native usage scenario, the default algorithm is single. When the cluster resources are insufficient, the service guarantee is generally achieved by the expansion of the cluster, and preemptive scheduling cannot be achieved during the scheduling process. Therefore, the original strategy cannot guarantee the operation of important services. Cloud video system provides video conference, online education, video office and other services. Cloud video consists of different pods like Nginx, MySQL, Tomcat, etc. This article sets pod priorities by dividing the categories of Cloud video Pods. Aiming at resource-constrained scenarios, the Preemptive resource scheduling strategy based on Cloud video Pod priority is proposed, and cluster experiments are carried out. Experiments show that the algorithm can effectively guarantee the operation of pods with high priority.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102R (2023) https://doi.org/10.1117/12.2671403
With the development of internet technology, more and more scholars are applying computer technology to research in the medical field. In this paper, we will investigate the named entity recognition method. Under the study of BERT-CRF model, we propose a named entity recognition model based on multi-task pre-training model with adversarial learning and network sharing and apply it to entity recognition in medical field with the aim of improving the accuracy of entity recognition in medical field. The model introduces multi-task joint learning and adversarial learning modules to improve the entity boundary effect and solve the noise problem of word boundary information, while achieving the purpose of information enhancement. On the CMeEE (Chinese Medical Entity Extraction) dataset, the model showed a significant improvement in accuracy, recall, and F1 score.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102S (2023) https://doi.org/10.1117/12.2671270
Accurately recognizing hand gestures has great significance in assisting human-computer interaction, enhancing user experience, and developing a human-centered ubiquitous system. Due to the inherent complexity of hand gestures, however, how to capture discriminant features of hand motions and build a gesture recognition model remains crucial. To this end, we herein propose a gesture recognition method based on multi-sensor information fusion. Specifically, we first use the accelerometer and surface electromyography (sEMG) sensor to capture the kinematic and physiological signals of hand motions. Afterward, we utilize the sliding window technique to segment the streaming sensor data and extract various features from each segment to return a feature vector. We then optimize a gesture recognition model with the feature vectors. Finally, comparative experiments are conducted on the collected dataset in terms of different machine learning models, different sensors, as well as different types of features. Results show the joint use of sEMG sensor and accelerometer achieves the average accuracy of 97.88% compared to the 90.38% of using sEMG sensor and 84.03% of using accelerometer among four classifiers, which indicates the effectiveness of multi-sensor fusion. Besides, we quantitatively investigate the impact of null gesture on a gesture recognizer.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102T (2023) https://doi.org/10.1117/12.2671064
Sarcasm is a special kind of linguistic sentiment that is widely used in a wide range of social media to express strong emotions in users. Therefore, the task of sarcasm recognition is particularly important for social media analysis. There are few studies on sarcasm sentiment recognition in Chinese, and they often ignore the complex interactions between different syntactic components of a sentence, such as sentiment words, entities, dummy words, and special punctuation that occur in the text. In order to improve the accuracy of Chinese sarcasm recognition, this paper proposes a multi-scale neural network sarcasm recognition algorithm incorporating a hierarchical representation of sentences, taking into account the semantic information of sentences and the relationship features between different syntactic components. The hierarchical syntactic tree is reconstructed to distinguish the key components of the sentence, and the multi-channel convolutional network is used to mine the relational features between syntactic levels and deeply fuse them with semantic information to perform the Chinese sarcastic sentiment recognition task. We have tested the method on a publicly available Chinese sarcastic comment dataset, and the results show that the method can effectively improve the accuracy rate of Chinese sarcastic sentiment recognition.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102U (2023) https://doi.org/10.1117/12.2671229
In order to reduce the incidence of traffic accidents, the use of computer vision to identify vehicles and passers-by in the process of driving can achieve the effect of assisting driving. This paper mainly introduces the performance improvement brought by the introduction of the SPP module in YOLO-V3 for object recognition. Model training is performed on the VOC dataset based on YOLO-V3-SPP. Finally, 300 photos were used to test the accuracy of the algorithm. The results show that the recognition accuracy of YOLO-V3-SPP for vehicles and pedestrians can reach 94.19% and 90.68%, and the accuracy of YOLO-V3 is improved by nearly ten under the same equipment. percentage point. The research on this technology can effectively reduce the probability of traffic accidents and provide reference value for the future driving safety warning field.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102V (2023) https://doi.org/10.1117/12.2671449
With the rapid development of 5G technology, the Intelligent and Connected Vehicle (ICV) technology is also evolving and expanding its application scenarios. In order to achieve lower latency and reduce the network load caused by massive data reflow in ICV, MEC (Mobile Edge Computing) technology is introduced to support ICV communication. While MEC technology brings a good experience to users, more and more attacks against Telematics come along, the most common of which is DDoS attacks, which can bring huge losses to telematics systems. Based on this, this paper proposes a DDoS attack detection method based on SAE neural network. The method uses the stacked Auto-encoder-based model proposed in the paper to detect network traffic in the telematics network, feeds the traffic data into the test model, and determines whether the automotive network system is under DDOS attack based on a threshold value. The DDoS attack is detected using the method proposed in the paper, with high detection rates in the training and test sets and stable models. Better experimental results were also obtained by later changing the number of hidden layers in the SAE network to detect DDoS attacks. Comparing the method in this paper with the SVM and CNN methods, the experimental results show that the DDoS attack detection method based on SAE networks works best.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102W (2023) https://doi.org/10.1117/12.2671063
With the development of the Internet, more and more service-oriented industries are transforming from stores to build platforms on the internet, and so is the second-hand car sales industry, which not only saves the cost of opening stores and employees, but also facilitates the majority of car enthusiasts. But changing the transaction model to O2O is even more technical and professional car issues transferred to the owners and buyers. Consider that some sellers or buyers will have inadequate preparation and thus suffer from the transaction. Therefore, the platform needs to give estimated prices based on previous normal transaction data and after confirming the owner's real submission of used car information, so that the owner can adjust between prices, thus ensuring the quality and speed of the transaction. In this paper, we used desensitized data on second-hand car transactions provided by WUBA. After data cleaning, the main model was constructed by neural network, and the model was trained with the processed data. After validating the model, the factors that affected the transaction cycle were found out and optimized the model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102X (2023) https://doi.org/10.1117/12.2671065
The depth face detector based on an anchor frame has demonstrated good performance with the advancement of science and technology, but tiny faces and partially occluded faces are still difficult to identify under the effect of posture, lighting, occlusion, and other factors. We present the ESSFD, an enhanced single-stage face detector, to increase the robustness of face detection in complex environments. The following are the important points: (1) Enhancing feature extraction by using an optimized ResNet-D network in the backbone network and adding Atrous Convolution Layer supplementary feature pyramids; (2) For data augmentation, use a combination of Gaussian blur and color jitter to lessen the impact of image environmental elements and increase the model’s robustness; (3) Fine-tune the training parameters for the model. In the presence of unusual postures, complicated lighting, and partial occlusion, ESSFD can increase the detection accuracy of tiny faces when compared to RetinaFace. Experiments reveal that in the easy and medium stages of the WIDER FACE dataset, the Average Accuracy (AP) of the ESSFD detector is around 1% higher than that of the advanced RetinaFace face detector (AP=95.373%). In the hard tiny face detection stage, it is around 2% higher than RetinaFace.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102Y (2023) https://doi.org/10.1117/12.2671062
Vehicles are closely related to human life, and the problem of road vehicles has always existed. In order to make road travel safer, the role of developing vehicle-road collaborative technology has become more and more important. However, there are still many problems in the monitoring of road vehicles, such as the integration of other single-point speed monitoring and video-based license plate recognition. In order to solve the problems related to vehicle speed monitoring, adapt to the current situation of insufficient utilization of road image data, this paper introduces a target detection algorithm based on YOLOv5, Dilb 's tracking algorithm, and perspective change speed solution. It is suitable for vehicle recognition tracking and speed calculation technology in general road camera perspective. The program designed by this technology has achieved high results in the actual test of the road. The mAP of the vehicle detection model after parameter adjustment can reach 94.7, and the speed measurement accuracy up to 90.3%, which has high usability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126102Z (2023) https://doi.org/10.1117/12.2671160
Clearance prices are vital for deciding on electricity spot market offers. To solve the problem of complex electricity assessment caused by frequent fluctuations in price levels, this paper investigates a data mining-based method for predicting the liquidation price in the electricity spot market. This paper establishes a data mining model to analyze and assess the forecast risk of electricity spot market clearing prices by solving for generalized forecast error parameters. On this basis, the correlation coefficients are solved to complete the design of the data mining-based power spot market liquidation price forecasting method. The experimental results show that under the action of the data mining model, the error level between the predicted and actual values of the electricity spot market clearing price does not exceed 200 yuan, which can effectively suppress the price fluctuation behavior and better solve the problem of complex electricity assessment.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261030 (2023) https://doi.org/10.1117/12.2672849
At present, the use of artificial intelligence in the front line of teaching has begun to take shape. Artificial Intelligence technology (AI) has triggered a new wave, and the innovation of science and technology has brought about the diversification of teaching methods. Fragmentary learning is everywhere all the time, online and offline are organically integrated, and knowledge breakpoints inside and outside the classroom are seamlessly linked up. Our classroom is changing with each passing day. This paper introduces AI technology to carry out the research of intelligent recording live broadcast of practical training, aiming at the problem of high picture stutter rate in the current process of recording live broadcast of practical training, which affects the subjective experience. Aiming at the practical training courseware with poor compatibility, the courseware is converted into H5 form courseware. On this basis, the data in the courseware is encoded and decoded. Through the real-time processing of guidance and control, the synchronization of video and audio recording, playing, and switching is realized. By using AI technology, live subtitle bars are automatically output. Through the comparative analysis of the application effect of the new recording live broadcast method, it is concluded that this method can effectively reduce the picture stutter rate and bring a better subjective experience to users.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261031 (2023) https://doi.org/10.1117/12.2671158
Serverless computing aims to handle all the system administration operations needed in cloud computing, thus, to provide a paradigm that greatly simplifies cloud programming. However, the security in serverless computing is regarded as an independent technology. The lack of security consideration in the initial design makes it difficult to handle the increasingly complicated attack scenario in serverless computing, especially for the vulnerabilities and backdoor based network attack. In this paper, we propose MDSC, a mimic defense enabled paradigm for serverless computing. Specifically, MDSC paradigm introduces Dynamic Heterogeneous Redundancy (DHR) structural model to serverless computing, and make fully use of features introduced by serverless computing to achieve an intrinsic security system with acceptable costs. We show the feasibility of MDSC paradigm by implementing a trial of MDSC paradigm based on Kubernetes and Knative. Analysis and experimental results show that MDSC paradigm can achieve high level security with acceptable cost.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261032 (2023) https://doi.org/10.1117/12.2680661
With the development of artificial intelligence technology, radio frequency identification technology is more and more widely used in wireless communication systems. This article describes an introduction to radio frequency identification technology. It also studies the design and application of wireless communication systems based on radio frequency identification technology.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Wenteng Liang, Shang Dai, Yizhen You, Kang Yang, Jianan Zhang, Tai Sun, Ruyi Li, Yue Zhang, Linxi Zou
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261033 (2023) https://doi.org/10.1117/12.2671453
In order to improve the accuracy of power dispatching text analysis and the ability to guide the operation of the power grid, a power dispatch text entity recognition method is proposed based on Bidirectional Encoder Representations from Transformers-Conditional Random Field (BERT-CRF). Taking the power grid fault handling plan text as the research object, the entity marking method of the fault handling plan is proposed. The word vector of the plan entity is calculated based on the BERT pre-training model, the characterization ability of the professional entity of the plan is enhanced by fine-tuning the initial BERT parameters, and the recognition ability of the plan text sequence is improved from the overall situation to access the CRF layer in the neural network. Thus, an entity recognition model of fault handling plan is established based on the BERT-CRF. Through the verification of a power grid fault handling plan, the proposed method has higher power dispatch entity and event recognition accuracy compared with other algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261034 (2023) https://doi.org/10.1117/12.2671044
In the direction of VR/AR human–machine interaction, natural and simple dynamic gesture recognition research has attracted much attention. For the sake of improve the accuracy of dynamic gesture recognition in Human–Machine Interaction, this paper proposes a new dynamic gesture recognition method FPN-3DResNeXt, which combines two-stream three-dimensional convolutional neural network (3DResNeXt) and feature pyramid (FPN). This method improves the structure of the 3DResNeXt network, adds feature pyramid and attention channel, optimizes the model parameters, and then improves the recognition accuracy; for the sake of improve the convergence speed and stability of the model, it is proposed to add batch normalization (BN) Further optimization of the network reduces the training time. The experimental results show that the dynamic gesture recognition rate of the method proposed in this paper is 95.30%, which is 2.1% higher than that of the gesture recognition method based on 3DResNeXt by comparing with various 3D convolution methods on the EgoGesture dataset, and it has better stability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261035 (2023) https://doi.org/10.1117/12.2671450
So as to reduce the interference of noisy information and enhance the accuracy of car marker image recognition in real life, a car marker image recognition model based on deep residual shrinkage network is proposed. The model combines attention mechanisms and soft thresholding function on the basis of deep residual network, which is used to eliminate noise and redundant information in the data, thus reducing the interference of noise information and improving the accuracy of image recognition. The experiments are conducted by adding Gaussian noise and pretzel noise to the car logo images shared by HFUT-VL on Github to simulate the car logo images captured under realistic conditions, forming a dataset of 200 images for each car logo with a total of 16000 images, and then training the deep residual shrinkage network with deep residual network and SENet algorithm model on the data. Then, the deep residual shrinkage network is compared with the deep residual network and SENet algorithm model to train the data, and some of the car mark images are tested. The results of the experiment show that the method has better performance than other deep neural network methods.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261036 (2023) https://doi.org/10.1117/12.2671184
In artificial intelligence, machines first need to be trained before they can work. Most of the time, they are trained on human data, such as human-made labels, or human physiological data. Data quality from humans becomes an important factor for machine learning. In the field of machine vision, machines usually learn human intentions through human gaze data, which is typically collected using video-based eye movement recording devices. Gaze recording accuracy can vary with pupil size, pupil size can be manipulated by screen background luminance. Looking at the variability of pupil size over various luminance levels may provide insight into possible ways to improve data quality. Aiming at improving the quality from human gaze data, we measured human pupil size changes during gaze at different luminance levels using a video-based eye tracker. Results show that mean pupil size gradually decreases, but is most variable at medium luminance levels, the maximum coefficient of variation difference reaches 7.5%. Therefore, gaze data accuracy may be improved if recorded at low or high luminance levels.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261037 (2023) https://doi.org/10.1117/12.2671122
Students' engagement is an essential indicator of students' participation in learning. This paper proposes an end-to-end three-stream fusion network students' engagement recognition model based on Spatio-temporal self-attention network. The model comprises three parallel TimeSformer networks, which extract facial information features, original video features, and video features after portrait segmentation. The calculation results of the three networks are fused for decision-making, and finally, the student's engagement level of the whole video is obtained. We trained and tested our model on the students' engagement dataset DAiSEE and obtained 57.8% four-level ACC. The recognition performance is significantly higher than the baseline of the database and better than other existing deep network models. The experimental results show that the accuracy of students' engagement recognition is effectively improved based on TimeSformer three-stream fusion network.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261038 (2023) https://doi.org/10.1117/12.2671094
According to the related M-matrix property, new upper bounds for the minimum eigenvalue of the irreducible M-matrix are provided. It is demonstrated that the new upper bound is sharper than the classical upper bound when the M-matrix is symmetric. Numerical examples further verify the validity of the results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Pei Yu, Pan Qi, Qian Wei, Lin Zhaoyu, Jin Xinyu, Chen Mingzhou
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261039 (2023) https://doi.org/10.1117/12.2671343
To develop rural e-commerce, we enhance sales of agricultural products by using speech recognition technology to lower the operation threshold. We also use a QR code-based traceability system to ensure product safety. We built a traceable e-commerce platform for agricultural items-based voice recognition.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103A (2023) https://doi.org/10.1117/12.2671073
The variable signal patterns and extremely high data rate of phased array radar greatly increase the complexity of electromagnetic environment, which makes the traditional method of radar working mode identification face great challenges. In this paper, a network structure based on temporal convolutional network (TCN) and Bi-directional long short-term memory (Bi-LSTM) parallel fusion processing is proposed. Depending on the advantages of TCN in depth temporal feature extraction and Bi-LSTM in global time series feature extraction, the typical working mode of phased array radar is accurately recognized. The experimental results show that under the condition of complex parameter interleaving, the recognition accuracy of the network for typical operating modes of phased array radar reaches 96.77%, which proves the feasibility of the method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103B (2023) https://doi.org/10.1117/12.2671438
In view of the large volume and complex structure of electromagnetic space big data, it is difficult to store and retrieve spectrum data using traditional databases and knowledge graph. Due to the abstractness and space-time characteristics of electromagnetic spectrum data, the use of event forms can better represent the spectrum data, and also make people and machines better understand. Based on the knowledge graph and the concept of events, this paper constructs the spectrum event knowledge graph (EMS-DEKG) and compares several methods of spectrum data retrieval through experiments, which shows that the EMS-DEKG method improves the stability and timeliness of electromagnetic space big data storage and retrieval.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103C (2023) https://doi.org/10.1117/12.2671079
In order to improve the timeliness of load feedback of streaming media cluster nodes and the processing efficiency of multiple concurrent requests, an improved algorithm based on dynamic feedback is proposed. The optimization of the method is as follows: (1) The calculation method of the load index weight coefficient and load weight value is improved, where the Least Squares is used to combine the optimization Analysis Hierarchy Process and Entropy Weight Method (denoted as LAE), which combine subjective weighting and objective weighting; (2) The feedback period is dynamically modified by the change in the number of tasks of the cluster nodes; (3) The Euclidean distance of the KNN algorithm is changed to a weighted Euclidean distance based on the weights obtained in (1). The cluster nodes are classified according to the improved KNN algorithm (denoted as LAE-KNN) and the load information, and the tasks are assigned to the class with the smallest total weight ratio; (4) At the same time, load migration is realized by setting the threshold of the load index, and random tasks of nodes exceeding the threshold of the load index are redirected to the low-load class according to the load information in each feedback cycle, to improve the load balancing effect of the cluster. Experiments show that the algorithm can effectively solve the problem of cluster load skew caused by many concurrent requests and can improve the load balancing effect of streaming media clusters.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103D (2023) https://doi.org/10.1117/12.2671200
Taking aerobics as an example, the human movement can be regarded as a series of posture data that changes over time. Compared with other methods, the special kinematic feature model of human skeleton has great advantages in describing the posture change state. In order to achieve the accurate capture of dynamic posture of aerobics, so as to complete the recognition and analysis of motion posture data in a short time, this paper proposes a 3D human dynamic posture recognition method based on Long Short-Term Memory (LSTM) network. First, the first frame model of the 3D human action sequence is selected as the template of the sequence, and the shape difference of the subsequent models of the action sequence is calculated by the shape difference operator relative to the template, which is represented as a low-dimensional shape difference information tensor. Then, the spatial and temporal dimensional features are extracted from the shape difference information tensor by combining two-dimensional convolutional neural network and LSTM to achieve the recognition of human dynamic posture. The above methods were evaluated by the dynamic pose datasets HumanEva, MoSh, SFU, SSM and Transitions; The classification accuracies were 98.4%, 99.7%, 100%, 99.4% and 100%, respectively.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103E (2023) https://doi.org/10.1117/12.2671164
Academic emotions are closely related to the achievement of online learning goals, for the current problem that discrete emotion recognition cannot describe the continuous dimensional academic emotion changes, this paper proposes a dimensional emotion recognition method based on a dual-streams convolutional feature ConvLSTM network. Firstly, static convolution features of single frame images in video sequences are extracted, and dynamic convolution features are extracted from video sequences, secondly, the concatenated double stream convolution features are analyzed by ConvLSTM network to extract the sequence features containing spatio-temporal information; finally, the generated feature passes through two full connection layers and outputs the predicted value of Valence-Arousal. The average of the Concordance Correlation Coefficient (CCC) on the public data set AVEC2015 reached 0.198. The experiment proved that the CCC correlation coefficient index of the proposed method on the AVEC2015 data set increased by 2.64%~6.28% compared with the baseline method, which can effectively identify dimensional emotions, providing a method support for dimensional emotion recognition.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103F (2023) https://doi.org/10.1117/12.2671521
With the rapid development of computer vision technology in the field of visual fashion, more and more people pay attention to the research on the "reliance" pattern of drama clothing. At present, in the field of clothing image display, research mainly focuses on clothing image recognition, key point detection, clothing recommendation, retrieval and matching. These studies can provide decision support for the design, production, display, sales and other links of drama costumes and bring a new display experience. However, in the realistic application scenarios of clothing "reliance" images, we still face challenges brought by changes in clothing style, materials, cutting, pattern composition and combination methods, which make the effects in recognition, positioning, recommendation and other applications continuously improved through experiments. The method based on depth learning in this paper focuses on clothing "reliance" pattern recognition, key point detection, clothing retrieval and other tasks.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103G (2023) https://doi.org/10.1117/12.2671180
The traditional tomato detection method of image segmentation is complex, and it is easy to be blocked by branches and leaves, fruit overlapping and other reasons, which affect the detection accuracy of fruit and the accurate positioning of picking points. This study proposes a fast identification and localization method of tomato based on YOLOv5 network. This method performs end-to-end detection by traversing the entire image with a single convolutional neural network, returning the class and location of the object. On the basis of YOLOv5, the regression box loss function is modified to improve the detection effect of tomato fruit, and the center point of the fruit boundary rectangle detected by YOLOv5 is used as the center point of tomato picking. The experimental results show that the average localization error of the proposed method is 1.379%, which is 1.867% lower than the traditional Hough method. The YOLOv5 method can effectively identify tomato fruits in natural environment. It can effectively detect tomatoes in overlapping, small targets, immature and other scenes, and perform more accurate positioning, laying a foundation for the tomato picking robot to select the best picking point.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103H (2023) https://doi.org/10.1117/12.2671154
With the rise and development of Internet of Things (IoT) technology, more and more devices access the network. However, most of them usually ignore the security issues, the crisis such as large-scale DDoS attacks caused by IoT botnet becomes more and more severe. It is significant to study the behavior of botnet and the detection technology. In order to improve the detection performance of IoT botnet, we analyze the behavior of botnet based on the traffic in IoT environment and propose a detection approach based on hierarchical clustering. Firstly, we capture the network traffic as .pcap files and aggregate packets into data flows based on five-tuple, then extract the basic statistics features by using a time window. Secondly, we analyze the typical features of IoT botnet during waiting period and malicious active period and optimize them by hierarchical clustering. Finally, XGBoost algorithm is used to classify the botnet. To demonstrate the effectiveness of the proposed approach, we trained KNN, Decision Tree, Random Forest models over the same datasets to detect IoT botnet and compared their performance with our approach. The experimental results prove our method can efficiently detect botnet as compared to other trained models.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qian Zhang, Yongzhi Zhu, Chuanhao Lan, Qinghang Mao, Yikai Cui
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103I (2023) https://doi.org/10.1117/12.2671057
In the era of information explosion, rumors will cause great harm and affect social stability. Most rumor detection methods concentrate on extracting features from content and consumer information. We propose a brand-new approach to early rumor identification, MSR-GAT. Firstly, the source text and comment text are fused as node features and the relation between events is considered edge information. Then, the graph attention model is constructed to classify nodes and complete rumor detection. The experimental findings demonstrate that the detection algorithm outperforms the baselines algorithm in accuracy, precision, recall and F1-Measure. It can accurately identify rumors.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103J (2023) https://doi.org/10.1117/12.2671666
The asphalt mixture mesoscopic model is mainly composed of aggregates, asphalt mortar, and the aggregate-asphalt interface layers. Based on Python, the built-in scripting language of ABAQUS finite element software, this paper provides a secondary development of the model to achieve modeling of random aggregate structure and asphalt mixture mesostructure. To solve the problem of aggregate interference, this paper compiles the vector operation of the GIS (Geographic Information Systems) algorithm with Python and applies it to eliminate the problem of aggregate location conflicts and overlaps. Based on the mesoscopic model, this paper simulates the splitting test of asphalt concrete by inserting cohesive units at the interface between aggregates and asphalt and between asphalt units, and the test results match with the actual test results and the simulation results of other scholars, which illustrates the reliability of the asphalt mixture mesoscopic model generated by this method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103K (2023) https://doi.org/10.1117/12.2671100
With the continuous progress of science and technology, architectural engineering is constantly diversified and intelligent, in a word, it is infinitely closer to more advanced science and technology. With the acceleration of urbanization, the number of apartment houses is increasing, which has become a symbol of modern cities. High-rise apartment buildings are characterized by many hazard sources, large fire load, high population density, complex building structure and difficult evacuation. Once a fire occurs, it will often cause huge economic losses and a large number of casualties. Performance-based design specifications give designers a lot of space, which is flexible and more reasonable. Because of these advantages of performance-based fire design, performance-based design has gradually become a method commonly used in building fires. In order to understand the smoke spread law and fire development characteristics of apartment building fire, this paper applies Pyrosim fire simulation software, through the establishment of fire model, set the corresponding parameters, can provide new means and tools for fire management, fire training, personnel evacuation, etc.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103L (2023) https://doi.org/10.1117/12.2671687
One-dimensional binary wavelet packet decomposition technology is applied to extract signal feature aim at the target acoustic echo signal in water of active sonar in this paper. The signal feature extraction method and the calculation of order list about decomposition node with frequency band are studied, which are verified and analyzed by different simulation experiment. All of studies designed to support a new technique and thought for the feature extraction of the target acoustic echo signal in water, which based on signal processing.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Shaojie Meng, Miaohua Huang, Yang Hou, Yanyong Ren
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103M (2023) https://doi.org/10.1117/12.2671163
Simultaneous Localization and Mapping (SLAM) research has become a hot spot in computer vision field. Static worlds are assumed in most existing SLAM systems. When traditional systems work in dynamic environment, the results of visual odometer will be affected by dynamic objects, resulting in low matching accuracy and poor performance. Aiming at the poor performance of traditional algorithms in dynamic environment, this paper proposes a new feature point extraction algorithm based on local feature matching relative geometry structure, which can efficiently extract robust features from image sequences in dynamic environment. Experimental results show that the features of dynamic objects can be correctly ignored by the proposed algorithm, and the feature matching accuracy can reach 98.8%, while maintaining the real-time property (the processing time of the same image with ORB is only 0.055ms longer).
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103N (2023) https://doi.org/10.1117/12.2671299
Binary code similarity is that different binary codes obtained from the same source code compiled by different compiler configurations are similar. Binary code similarity detection is often used to evaluate whether functions in two binary codes are similar. This technique has critical applications in intellectual property protection and IoT security, such as code plagiarism detection, malware detection, vulnerability detection, etc. In this paper, we propose a text semantics-based binary function similarity detection model SBFS, which firstly transforms binary functions into function texts by preprocessing assembly instructions; then learns function texts to obtain semantic embedding vectors using a natural language processing model. Finally, the similarity between two functions is measured by calculating the cosine distance between the embedding vectors of the two functions. Experimental results show that the SBFS model can achieve cross architecture detection and higher accuracy with 98.2% in the binary function similarity detection task.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103O (2023) https://doi.org/10.1117/12.2671213
Through the rapid quantitative loading system, a large amount of coal bulk material is loaded into train or car carriages. In the process of coal loading, the eccentric loading problem often occurs because of the uneven distribution of accumulation density. The discrete element method is applied to establish the simulation model of the coal loading system, and the complex particle model with different shapes is introduced. The accumulation characteristics of bulk material at different moments during the coal loading process are analyzed, and the accumulation distribution characteristics of different types of particles are analyzed. The hierarchical distribution law of coal bulk layered accumulation density in the carriage is obtained, which provides a theoretical basis for the uniform loading of coal bulk.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qin Zhong, Chunyan Zhao, Xin Zhou, Yan Wang, Ling Li
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103P (2023) https://doi.org/10.1117/12.2671104
For the Hadamard product of the matrices with non-negative entries, we study the new upper bound for the spectral radius by applying the characteristic value containing the domain theorem. This estimating formula only involves the entries of two non-negative matrices. Hence, the upper bound is easy to calculate in practical examples. An example is considered to illustrate our results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103Q (2023) https://doi.org/10.1117/12.2671383
When designing international student management information system, we should fully consider the needs of the users. International students have a short time to learn Chinese, especially when it comes to some of the proper nouns, they are often prone to ambiguity, so we can consider the multinational language requirements when designing the management system, so that international students can understand the relevant issues through their own language when they encounter problems. Therefore, this paper is based on cloud computing technology, the current popular Java EE development platform, and the mature MySQL database to design the academic management software. First, this paper analyzes cloud computing and the architecture used in this paper. Then, the overall structure of the system is given in this paper, taking into account the training needs of international students. Then a detailed design is given for each part. The designed system is also tested, and the test results show that the performance as well as the functionality of the system can meet the current requirements.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103R (2023) https://doi.org/10.1117/12.2671096
Unmanned equipment is more and more widely used in the battlefield. The interface types of photoelectric and radar equipment equipped on unmanned reconnaissance ships, unmanned surface vessel and unmanned chariots are complex, and the integrated video bandwidth is large. Although the navigation and cluster capabilities of unmanned equipment are becoming more and more intelligent, human-computer interaction is still very necessary in key steps such as real-time battlefield situation analysis and attack target confirmation. The communication bandwidth between unmanned equipment and control center is limited, which is seriously contradictory to the high bandwidth of sensor videos. The design of video co-processing system is based on Hi3531 processor and TMS320C6678 DSP, which may adapt to photoelectric tracker, analog search radar, network navigation radar, optical fiber interface multi beam radar videos, and accomplish the video compression and decompression processing to reduce the bandwidth requirement for communication system. On the premise of ensuring real-time video transmission and reducing video transmission bandwidth, the video display effect is optimized. The oscilloscope test shows that the compression and decompression delay of TV is 16.3ms, and that of radar is 19.5ms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103S (2023) https://doi.org/10.1117/12.2672723
The current traditional power grid marketing cost forecasting method achieves cost forecasting by studying relevant project examples, which leads to poor forecasting results due to the shallow analysis of cost influencing factors. In this regard, an improved ARIMA model-based power grid marketing cost forecasting method is proposed. The key factors affecting the cost of grid marketing are analyzed by using principal component analysis, and the main indicators affecting the cost are selected by comparing the degree of influence of each indicator on the overall, and the cost prediction model is constructed by using ARIMA algorithm. In experiments, the proposed cost prediction method is validated. The prediction results of the proposed method are in good agreement with the actual cost situation, and the cost prediction performance is better.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Junpeng Dang, Jinhai Yang, Xuefeng Liu, Lei Yu, Ying Lin
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103T (2023) https://doi.org/10.1117/12.2671351
Aiming at the problem that the coordinates of the existing fault point detection methods are inaccurate when determining the specific location of the fault point, this paper introduces the fault tree to design and study the over-current fault point detection method of the transmission cable ring network. For the normal and abnormal operating states of the transmission cable, determine the transmission cable's farthest over-current signal flip point fault string. The fault tree is introduced, and the over-current fault of the transmission cable ring network is substituted into the fault tree structure to realize the specific description of its characteristics; Finally, the particular location of the fault point is determined by rough measurement and precise measurement, and it is imported into the standard coordinates to realize the fault point detection. Compared with other detection methods, the positioning results obtained by the new detection method are more realistic, the detection accuracy is higher, and the specific location of the fault can be accurately found.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103U (2023) https://doi.org/10.1117/12.2672738
With the continuous development and maturity of information technology, fire protection and security have always been the key issues that perplex the safety construction of closed places. In the face of the large demand for sensor equipment, part of the low-carbon park fire safety monitoring system has the problem that the corresponding time is too long. A low-carbon park fire safety monitoring system based on a multi-sensor is designed. The liquid crystal display screen of the LCD 1602 is selected as a display device, and the standard 16-pin interface is expanded to 18. It collects the fire and security linkage data of the low-carbon park, compresses the data volume on the basis of homogenization processing of the original data, and quickly confirms the location information according to the per-implanted electronic map. It then delineates the dangerous area and optimizes the alarm function of the monitoring system by using multiple sensors. The experimental results indicate that the average response time of the proposed system is 13.416s, which is 6.261s and 6.587s lower than the other two systems, indicating that it is more suitable for the actual scene after multi-sensor fusion. At the same time, it also enriches the theory and practice content of the monitoring system research field.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103V (2023) https://doi.org/10.1117/12.2671091
Medical review on the medical platform contains valuable information, including patient concerns, opinions, suggestions and emotional tendencies. This paper utilizes text mining technology to investigate consumers' emotional tendencies, automatically identify keywords in Internet hospital review texts, and use theme models, word embeddings and other technologies to analyze consumers' potential needs and wishes. The research results can provide an important reference for smart medical port managers to formulate targeted management and promotion strategies.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103W (2023) https://doi.org/10.1117/12.2671314
This design first carries on the detailed background analysis to the patient monitoring system and explains the main technology and technical difficulties of the system development. After a deeper analysis of the patient monitoring system, the use of Java as the most important development tool, mainly applied to the development and network retrieval, MySQL as the database of patient monitoring, JSP through the web page for data interaction, complete data access, and then create a web page dynamically. After the preliminary design of the patient monitoring system, the software debugging is carried out to ensure the safety and efficiency of the system. The results show that patients and doctors can use the system to find the information they want.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103X (2023) https://doi.org/10.1117/12.2671701
At present, Chinese electrical fire prevention and control is still in the development stage. The existing electrical fire monitoring products have technical defects, the false alarm rate is high, and the monitoring effect on electrical fire is limited. Because of these situations, the microcontroller STM32WLE5CCU6 and the sharp micro power quality chip RN8302B is used as the core, combined with the LSTM neural network, an intelligent front-end device with real-time monitoring of electrical fire data and data depth processing functions are designed, which realizes the monitoring and prediction of an electrical fire. And composed an intelligent electrical fire warning system capable of wireless communication and information interaction with intelligent front-end equipment, LoRa wireless communication, RS485 communication, and cloud platform.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103Y (2023) https://doi.org/10.1117/12.2672753
In the wake of developments in science and technology, wireless sensor networks have attracted more and more researchers. Compared with traditional networks, wireless sensor networks are very different in many respects. Energy saving, security and reliability are the most concerned topics of wireless sensor networks. As the network expanded scale and complexity are increasing, the limited sensor node resources and the uncertainty of the environment layout, the traditional security mechanism is no longer applicable to wireless sensor networks. This article reviews the common assault forms, security technologies and security mechanisms of trust-based routing protocols proposed in wireless sensor networks in recent years. Based on the analysis of various types of attacks, the development direction of the trust mechanisms in wireless sensor networks is provided.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126103Z (2023) https://doi.org/10.1117/12.2671366
Colorectal cancer is one of the most common malignancies that can develop from high-risk colon polyps. Currently, the application of deep learning to colon polyp segmentation has obtained high accuracy, which can have good detection rate for small polyps, but due to the limitation to the morphological diversity of polyps and unclear boundary between polyps and surrounding mucosa, there will be insufficient accuracy and unclear segmentation of marginal regions. To address the above problems, we propose Shifted Windows Parallel Network (SWPNet) for real-time accurate polyp segmentation. We design a four-layer Swin-Transformer module with window segmentation as the backbone network, which have strong feature extraction capability, and on top of the improved version of U-Net structured network with additional encoders and decoders, we add Channel Feature Pyramid (CFP) blocks on the deepest three layers of feature extraction, set different rates of dilated convolution respectively, and fuse more scales of semantic information to refine edge segmentation. We conducted experiments on each of the five classical polyp segmentation datasets and achieved high accuracy, especially on the challenging ETIS dataset with 74.1% mean Dice and 65.6% mean IoU respectively.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261040 (2023) https://doi.org/10.1117/12.2671189
With the increasing scale of high-voltage cable equipment in domestic urban power grids, it is necessary to deepen the intelligent construction of transmission lines, solve the common problems encountered in big data processing and edge side application of high-voltage cables, and take edge IOT agent as the cutting point for technical research. By studying edge computing, AI image recognition and intelligent linkage control model of cable channel business application, intelligent management and control of high-voltage cable line status, risk early warning, differentiated operation and maintenance decision, etc. can be realized, and the intrinsic safety level and lean operation and maintenance management ability of cable lines and channel equipment can be improved.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Chongxiao Qu, Yongjin Zhang, Yufeng Wang, Changjun Fan, Shuo Liu
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261041 (2023) https://doi.org/10.1117/12.2671221
With the rapid development of information technology, the volume of data proliferates, and so does the increasing demand for data reliability. As a result, high-availability network storage systems have been paid more and more attention. In view of the above situation, considering the current research status and development trends in this field, this paper proposes a design and implementation scheme of a high-availability system for storage networks based on PCI Express (PCIe) non-transparent bridge technology and applies it to a dual-controller storage array. It focuses on the design of high-availability network storage and standard network interface based on non-transparent bridges. Finally, through experimental validation and analysis, our solution can accomplish the functions of storage fault handling and I/O forwarding while having the advantages of high bandwidth performance, high availability and high scalability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261042 (2023) https://doi.org/10.1117/12.2671265
Along with social development, the problems of insufficient rural labor and mismatch between labor intensity and economic returns have become a greater obstacle to rural development; therefore, the development of agricultural intelligence to improve agricultural productivity will become the new direction of modern agricultural development. Based on internet technology, intelligent agriculture adopts digital technologies such as intelligent perception, network transmission and big data processing to provide decision basis for agricultural planting, production and pest control, or directly deploy decision information to automated farm equipment, so as to use agricultural resources reasonably and efficiently. However, the development of smart agriculture is constrained by the low accuracy of automation control and decision information. In this paper, we optimize the threshold value of each agricultural production decision data through machine learning algorithm, use Bayesian optimization to learn the agricultural production environment and crop growth data, iteratively optimize the threshold value to get the global best value, improve the accuracy of automation control and reduce the risk of agricultural production decision, and effectively improve the economic returns of agriculture.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261043 (2023) https://doi.org/10.1117/12.2671330
Flight test comment is one of the important links in flight test engineering, which plays an important role in summarizing the flight situation, analyzing and locating the fault problems, and arranging follow-up tasks. Aiming at the problems existing in the traditional flight test comment system, such as single evaluation process and method, insufficient data support, insufficient visualization and interactivity, and unsatisfactory effect and efficiency, this paper constructs a digital intelligent flight test comment system based on data mining and software technology. It mainly realizes the functions of basic task information display, key data and scene reconstruction, fault analysis and location, and automatic generation of task evaluation result report, so as to realize the digitalization and intellectualization of the future flight test evaluation business.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Fang-ju Ran, Chen-zhi Xiong, Meng-yao Lu, Tian-qing Yang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261044 (2023) https://doi.org/10.1117/12.2671058
This paper proposes a method of emotion analysis based on BERT BiLSTM. Firstly, BERT is used to realize the word vectorization, and then Bilstm is constructed to extract semantic features for emotional analysis. In the experiment, the model designed in this paper is compared with the emotional dictionary, SVM, Word2vec LSTM, BERT TextCNN on the college online public opinion comment dataset, and the experiment proves that the accuracy of this model has been improved.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261045 (2023) https://doi.org/10.1117/12.2671114
The purpose of this design is to improve the efficiency of cupping-induced plaque sorting and to meet the needs of the people to complete cupping-induced plaque sorting accurately. Deep learning methods and image classification techniques are used in the classification system to construct a classification model for cupping-induced plaques by migration learning on the RESNET convolutional neural network. The system classifies the cupping-induced plaque images based on that uploaded by the population. The experimental results shows that the classification accuracy of each category is above 80%, which verifies the effectiveness of the classification system.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261046 (2023) https://doi.org/10.1117/12.2671068
With the impact of the epidemic, enterprise risk assessment has gradually become hot research. At present, enterprise risk assessment is still based on financial data, and there are few studies focusing on text type data. However, text type data has a certain impact on enterprise risks and will affect the overall development trend of enterprises. Therefore, how to build the enterprise risk assessment model and judge the current risk situation according to the text data is the main challenge faced in the current field of enterprise risk assessment. Based on the above problems, the enterprise risk assessment system based on text mining is designed to realize the enterprise risk trend judgment. Focus on the five aspects of the enterprise risk assessment system: first, obtain the corresponding evaluation data according to the enterprise risk assessment task, and build the corresponding database; second, build the enterprise risk assessment field knowledge base to provide knowledge support for the subsequent enterprise risk assessment; third, put forward the named entity identification algorithm for enterprise risk assessment; fourth, propose the relationship extraction algorithm for enterprise risk assessment; fifth, design the enterprise risk assessment model to realize the enterprise risk tendency judgment. Finally, the availability of the enterprise risk assessment system is verified based on the text type data of a particular enterprise.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261047 (2023) https://doi.org/10.1117/12.2671098
In order to solve the problem of unbalanced Session Initiation Protocol (SIP) requests distribution in SIP server clusters when processing concurrent task requests, an improved dynamic load balancing optimization algorithm based on SIP transactions is proposed. Firstly, according to the characteristics of SIP protocol, SIP transactions are used as the unit to measure the load, different weights are assigned to different types of transactions, and the variability between server nodes and network performance are considered to measure the load of server nodes together with the response time. Then, the load factor values dynamically obtained are used to calculate the load ratio of the server nodes and compare it with the set threshold value, and the server node with the lowest load ratio in the normal state is selected to allocate SIP requests according to the comparison results, which allows the server nodes to have a buffering process, reduces the occurrence of throughput jitter phenomenon, and effectively avoids system overload. By using the open-source testing tool SIPp to conduct experiments, the results show that the improved method can solve the load imbalance problem, which makes the SIP server cluster system allocate resources efficiently and reasonably, improves the utilization of system resources, has a better balance and reliability compared with other algorithm models, and achieves the expected results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261048 (2023) https://doi.org/10.1117/12.2671143
Task allocation is a significant topic in the field of artificial intelligence, and ideas for solving this problem are often drawn from the widespread phenomenon of division of labor in nature. For the problem that the existing labor division game model does not match the characteristics of the large-scale division of labor scenarios, we propose a new model named Division of Labor within Community game (DOLC). The article establishes the mathematical model of DOLC, then we investigate the effects of selection intensity, urgency coefficient and responsibility coefficient on the evolution of division of labor. We derived the following conclusions: When both urgency and responsibility coefficients are zero, this causes the tragedy of DOLC, while group has the opportunity to maintain the division of labor if one of them is positive. The responsible group has always been cooperative. Relative to urgency coefficient, responsibility coefficient plays a more noticeable role in facilitating the evolution of cooperation in population. DOLC model is well suited to task allocation problem in multi-agent systems, and the findings of this article provide new insights for improving the intelligence of unmanned systems and optimizing task allocation and scheduling.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261049 (2023) https://doi.org/10.1117/12.2671086
This paper mainly studies the classification and detection of fruit based on deep learning. In order to solve the problem of current industrial demand, this paper proposes a new method for classification and detection of fruit based on YOLO. In the detection algorithm, the target detection and classification tasks are treated as a regression problem. It is convenient to transplant to different devices in the complex environment of industrial environment. A multi-feature fusion and multiple pre-selection frame prediction strategies to solve a variety of fruit classification problems. The test results show that the system can achieve the accuracy, real-time and compatibility required by the industry.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104A (2023) https://doi.org/10.1117/12.2671432
It is well known that the reflection of mirrors can provide spatial perception, and this visual experience can increase immersion in VR. However, it is a challenge to integrate the mirror element, where reflection is the only feature, into an interactive VR game. This paper is a new attempt to put multiple mirror puzzles into VR environments using four game mechanics of mirrors, namely, single reflection, recursion, light path change, and duplicated spacing. Three major levels were designed using these mechanics, the Mirror Maze, the Laser Puzzle, and the Key to Yin and Yang. The findings of the game experience demonstrate that the combination of audio-visual interaction increases the immersion of the game, the easy-to-understand puzzle solutions meet the comfort level of the player experience, and the combination of east-west elements enriches the quality of the game. Overall, this is a game design attempt worthy of being studied, and it is feasible to disperse a single feature of the mirror into multiple mechanisms to add to the player experience for an interactive game.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104B (2023) https://doi.org/10.1117/12.2672730
Liquid level monitoring is widely used in industrial systems monitoring. Accurate monitoring of liquid levels is an essential tool in the production control process, especially for monitoring the status of oil storage tanks. In order to improve the accuracy of the storage tank monitoring system, this paper designs an intelligent monitoring system based on infrared thermal imaging for storage tank level monitoring. The hardware part of the level measurement system is mainly composed of the host computer, microcontroller, head, and infrared thermal imager. The software design consists of the following aspects: level image acquisition based on infrared thermal imaging, level measurement using ultrasonic characteristics, and level monitoring using input transmitters. By comparing this system with a conventional monitoring system, the experimental results were obtained: the relative error of the monitoring system in this paper was 0.8%. The relative error range of the conventional system is 2.6% to 5.8%. It can be seen that the system in this paper is better than the traditional system, and the experiment is successful. The oil storage tank level monitoring system designed in this paper has the advantages of accurate monitoring, stable operation, simple operation, and simple implementation of network management, and it has broad application prospects.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104C (2023) https://doi.org/10.1117/12.2671447
The coolant liquid nitrogen, which is commonly used in medical treatment at present, is prone to frostbite and vaporization, and the greenhouse effect of Freon is serious. With the high-speed development of CO2 liquid storage technology, we design an STM32-based control system for cryogenic shock therapy using CO2 as coolant. The control system uses STM32F103RCT6 microcontroller as the core processor with modular design, including main control module, infrared temperature measurement module, human-computer interaction module, serial communication module, control output module, power supply module and so on. The modules work together to realize the functions of serial communication, real-time temperature monitoring, solenoid valve start-stop control, etc. The distance between the nozzle and the cooling substrate and the effective area of cooling are studied quantitatively by experiment under the straight tube type nozzle with 0.5mm inner diameter and 4cm length model. The results show that the control system is stable and practical, and the ideal distance between the 0.5mm*4cm nozzle and the cooling substrate is 30mm for the best cooling effect, and the effective range is the ellipse area of 1.8cm for the long half-axis and 0.9cm for the short half-axis, which provides a reference for clinicians to perform spray treatment operation on the damaged tissue epidermis of patients.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104D (2023) https://doi.org/10.1117/12.2671099
When an abnormal flight occurs, if the previous flight cannot take off as planned, it will affect the subsequent flight, resulting in a downward impact. Therefore, airlines often adopt different recovery measures (including flight delays, flight cancellations, aircraft swaps, etc.) to eliminate or mitigate the downward impact. When evaluating the pros and cons of the recovery plan, the loss of delay, loss of flight cancellation and loss of aircraft exchange are generally considered. However, in fact, many complex factors are ignored when measuring these losses, such as food, transportation and accommodation costs of crew and passengers caused by flight delay, and compensation for delay, etc. Expert systems are suitable for situations where no or little data is available and the business logic is complex, and their introduction into flight disruption impact assessment is an exploration of artificial intelligence in civil aviation. The evaluation of the impact of flight disruptions by an expert system not only quantifies the benefits of recovery solutions, but also provides some reference for evaluating the advantages and disadvantages of existing models and algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104E (2023) https://doi.org/10.1117/12.2671175
The tongue organ has been widely studied by researchers because of its special status in Traditional Chinese Medicine (TCM). Doctors can determine the health condition of patients by observing their tongues. With the development of Artificial Intelligence (AI) technology, many efficient models have been designed for automated diagnosis, especially for mobile devices. In this way, many people can use the tongue diagnostic system deployed to mobile devices for health management. However, the small memory and low imaging quality of mobile devices have been limiting the efficiency of diagnosis. To address these problems, we design the efficient and accurate network for real-time tongue segmentation (RTS-Net) on mobile devices. The RTS-Net consists of two parts: lightweight ghost encoder for accurate feature extractor with less parameters and efficient decoder to recover the tongue details. Specially, we take GhostNet as backbone and remove its last avgpool layer, fully connection layer and pointwise convolution. Then, the ASPP module is adopted to capture multi-scale features and abstract semantics. We also design an efficient and accurate decoder to recover the resolution of features as well as compensate for feature details. We collect tongue images taken from mobile devices on the web and build the corresponding dataset to test the effectiveness of our model for low-quality images taken by mobile devices. Overall, the dataset contains images of tongues taken in different backgrounds, various angles, diverse ages and non-uniform lighting. Extensive experiments are conducted on our mobile tongue dataset and the result shows that proposed method is lightweight and accurate for mobile tongue segmentation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Changjun Zhao, Ming Cheng, Wei Chen, Jianyu Wu, Jieyun Xie, Juandi Li
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104F (2023) https://doi.org/10.1117/12.2672734
Aiming at the problem that the answer given by the existing power marketing customer service system does not match the standard answer in customer question-and-answer, this paper introduces knowledge mapping to carry out personalized design research of power marketing intelligent customer service system. A framework structure is designed according to the functional requirements of a customer system. The selection of a database server, an application server, and a mobile monitoring terminal device is completed based on the hardware requirements of the system. A power marketing knowledge base is constructed by using a knowledge map. Finally, personalized customer service intelligent question answering is realized through Chinese processing and label scoring. Through comparison and verification, the new customer service system can give reasonable and accurate answers according to customer questions in power marketing, and effectively promote the service level of power marketing.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104G (2023) https://doi.org/10.1117/12.2671287
As economic globalization advances, financial market is increasingly favored by investors. With the development and strong demand of financial market, the forecast of stock price trend has aroused widespread attentions from both the academic and industry. As is well known, stock investment has both high returns and high risks. However, it is difficult to quantify the internal and external factors that affect stock market fluctuations, and it is also difficult to process massive and complex stock data. Therefore, traditional non-artificial intelligence approaches are not always satisfactory in forecasting stock price. Therefore, it has great significance to use big data technologies to excavate massive useful information hidden in stocks as well as to use neural network technology such as LSTM to further solve the problem of stock price trend forecast. In the paper, we report a development and implementation of deep learning-based stock opening price forecasting system based.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104H (2023) https://doi.org/10.1117/12.2671085
Angle of arrival-based localization has been widely applied in wireless sensor network due to its easy access of measurements and simple system structure. In previous work, the positions of anchor nodes are generally assumed precisely known, which is an important condition in addition to the angle of arrival measurements. However, in practice, anchor positions are obtained by global positioning system or other localization methods, which inevitably suffer from errors. These errors, leading to inaccurate anchor positions, have heavy impact on the localization result. Although there are a few studies on the errors of anchor position in wireless localization, they mainly focus on range-based measurements. In this paper, we study angle of arrival-based localization with inaccurate anchor positions. Stemming from the maximum likelihood estimation, a novel semidefinite programming method is proposed by using tight approximation and proper relaxation. Numerical examples demonstrate that the proposed method provides much better performance in terms of localization accuracy, compared to some existing methods.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104I (2023) https://doi.org/10.1117/12.2671359
Deep Neuron Networks (DNNs) rapidly develop across numerous industries and fields. The Residual Network (ResNet) is prevalent for its high performance and novel residual network for image classification. One new architecture derived from ResNet, the ResNet-RS, is proving to have higher accuracy while having lower training difficulty. Furthermore, with the growing network depth and expanding training set size, it is vital to apply different parallel methods to the ResNet to improve the training speed. To evaluate the parallel efficiency of the new architecture, this paper made the following efforts: (1) Implemented the ResNet-RS network on VPipe with a two GPU environment, (2) proposed a formal procedure for DNN implementation on VPipe, and (3) compared the top one and top five accuracy, loss, and epoch time of training between an orthodox ResNet and the derived ResNet-RS on VPipe parallel system.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104J (2023) https://doi.org/10.1117/12.2671719
Contemporarily, plenty of milestones of black hole in the field of astrophysics have been achieved, e.g., discovery and observation of the black holes. On this basis, this paper summarizes the recent progress of relevant studies from the simulations and state-of-art detectors. A combination of quantitative and qualitative approaches was utilized for data analysis. Specifically, we explore the laws of black hole mechanics and its resemblance to the laws of thermodynamics, which leads to one of the most important works of black hole, i.e., the Hawking radiation. Besides, the category of different black holes is introduced in terms of angular momentum and electric charge. Moreover, we summarize some observed data and demonstrate the meaning of collected photograph with theoretical analysis of black hole observation and discuss some observed law in the different spectrum (e.g., radio loudness and energy spectral slope). These results shed light on further investigation for black holes.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104K (2023) https://doi.org/10.1117/12.2671187
Previous Multi-Agent Path Finding (MAPF) solvers rely on several simplifying assumptions. They consider the agents as holonomic robots and ignore the agent’s size. There are two main directions of MAPF research. One is about MAPF’s own shortcoming, how to improve the efficiency of the existing algorithm and the quality of the solution. The other direction is how to deal with diverse constraints brought by different cases when applying MAPF to practical problems. In this paper, the latter direction for solving the handling problem of forklifts in the roadway is more focused on. Firstly, we define MAPF for forklift robot which considers the agent’s size and the topological paths as vertices. Secondly, we exploit the advantages of ECBS to make PBS runs faster with a bounded suboptimal solution. Then, we propose an acceleration strategy for collision detection, which is an important factor limiting the speed of the entire algorithm. Finally, we use the weighted low-level search to speed up calculation. The above series of improvements have enabled the MAPF algorithm to be applied in our real scenarios.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104L (2023) https://doi.org/10.1117/12.2671209
Chinese pattern design has not only witnessed the history of more than 5,000 years in China, but also impacted the aesthetic cognition of the East in the West. Chinese patterns have been beautiful since ancient times. In the past, it was created by the wisdom and hard work of the Chinese people, and now it should be inherited by the wisdom and hard work of the Chinese people. For the visualization platform of Chinese pattern design, better construction and improvement are needed. Therefore, in order to visualize information and achieve better results, if the memory occupancy is too high, the operation effect of the platform will be reduced. In order to improve the operation effect of the visualization platform, the construction of visual virtual reality platform of Chinese pattern design is proposed. Based on B/S mode, the software structure is established, and specific analysis is carried out, and the functional plate and visual effect design are improved. Through hardware and software design, the visual virtual reality platform of Chinese pattern design is constructed.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104M (2023) https://doi.org/10.1117/12.2671263
In this paper, a multi-state AIDS model is developed and used to analysis the potential impact of a non-complete vaccine immunization. The model is a nonlinear ODE system, also a differential dynamical system. In this paper, we divide AIDS into multiple-state and construct ODE for the change in the number of people in each stage. The main point of this paper is to proof that the transmission of AIDS in the population is stable under vaccine immunization. The form of stability under different conditions is also given. Firstly, our article describes the parameters and builds the model, then verifies the well-definition and positive invariance of the model. Next, the concept of basic reproduction number ℛ0 is introduced from the next generation matrix. Importantly, in order to verify the disease-free equilibrium of the model, we use the Lyapunov function to proof globally asymptotically stable of the ODE system. In the proof, we know: the system is globally asymptotically stable when ℛ0 ≤ 1. In addition, when ℛ0 ⪆ 1, the system has a unique endemic equilibrium solution, i.e. locally stable. Finally, after the proof, we conclude: (1) If we can keep the ℛ0 ≤ 1, AIDS will die out when time tends to infinity. (2) If ℛ0 ⪆ 1, the infected population would exist in the population at a constant rate for a long time, that is, the proportion of the infected population tends to 1 - 1/ℛ0. The main innovation of this paper is to use a system of ODEs to describe the infection status of the AIDS population.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104N (2023) https://doi.org/10.1117/12.2672192
In order to solve the problem of sparse and missing example in the off-line handwritten signature authentication scenario, a handwritten signature authentication tree algorithm was proposed. The algorithm improves the perceptual hash algorithm based on the discrete Fourier transform, which is used to gather the subject information in the signature image, compress and generate the perceptual digest; then the cosine distance is used to calculate the distance between the perceptual digest vectors. According to the cosine distance matrix, an authentication tree composed by multiple nodes is constructed and the distance threshold is determined to authenticate the unknown data. The authentication result can be used for the self-renewal of the authentication tree. The experimental results show that the handwritten signature authentication tree algorithm has high accuracy and strong robustness. The false rejection rate and false acceptance rate on small datasets are significantly lower than the traditional machine learning algorithm, and its self-renewal mechanism can also cope with the style changes of signature handwriting very well.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Yongming Liu, Qiang Ma, Lei Fu, Zhuanzhe Zhao, Zhibo Liu, Zhijian Tu
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104O (2023) https://doi.org/10.1117/12.2671243
Due to the confidentiality of RV reducer's performance indicators and other issues, it is difficult to obtain its technical data and bring some difficulties to its theoretical calculation. At the same time, the performance indicators calculated theoretically are often limited by various factors and can only be used as reference values. Transmission efficiency is one of the most important performance indicators of RV reducer. Therefore, this paper takes the transmission efficiency of RV reducer as an example, aiming at the problem that the coaxiality error of the traditional RV reducer testing system is lack of compensation and adjustment, based on the error compensation principle, a coaxiality error compensation mechanism is proposed to adjust the coaxiality error of the RV reducer to reach the allowable range. Finally, the prototype experiment shows that this mechanism can realize the coaxiality error compensation, and the average transmission efficiency in the steady state is 31.98%, which verifies the accuracy of the compensation mechanism. This research has certain theoretical significance and practical value for improving the performance detection accuracy of RV reducer.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104P (2023) https://doi.org/10.1117/12.2672848
The Black-Scholes (B-S) and Leland models play an important role in option pricing. This paper investigates the differences between the Black-Scholes model and the Leland model from aspects of option pricing in real markets with or without transaction costs. Specifically, the results of the B-S models and Leland model performance differently when transaction costs are introduced. The paper demonstrates the differences between the two models by selecting the relevant prices of four stocks for the last ten days and calculating their theoretical prices based on the relevant algorithms of the two models. According to the analysis, every change on conditions has the potential to affect the pricing of options in the market. Based on these results, we have focused on the differences in option pricing after the implementation of transaction costs, though some realistic factors that may affect the results have been ignored in the calculation process for both models. Overall, the greater the transaction costs, the larger the deviation from the analytical results of the two models. These results shed light on option pricing in a real market.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Wei Xie, Haihong Du, Wenming Pan, Helong Wang, Hongxin Liu, Mingxing Zhang, Jun Wang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104Q (2023) https://doi.org/10.1117/12.2671310
With the promulgation of the notice of market-oriented reform of coal-fired power generation, the reform of transmission and distribution electricity price has entered a substantive reform stage. General industry and Commerce have all entered the market, the proportion of electricity marketization has increased significantly, and the profit space of power grid companies has been further narrowed. Facing the dual pressure of performance assessment of state-owned assets management commission and verification of transmission and distribution electricity price, investment, as an important driver of business growth, the scientific value of its front-end decision-making has an important impact on stabilizing transmission and distribution electricity price and improving the company's business performance. However, the current investment plan arrangement of front-end infrastructure projects is not effectively connected with the key elements of transmission and distribution electricity price verification, and there is still great uncertainty about the supporting role of transmission and distribution electricity price. There is a lack of overall arrangement in advance, which is not conducive to improving the company's operating performance. Therefore, this paper excavates the investment and asset formation laws of infrastructure projects, makes the linkage relationship between investment, asset formation rate, transmission and distribution price and business performance smoother, studies and builds a power grid investment decision-making model based on the constraints of key elements of transmission and distribution price, provides quantitative reference for investment arrangement, and improves the scientific value and rationality of investment arrangement at the source.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104R (2023) https://doi.org/10.1117/12.2671514
Since the outbreak of COVID-19, it has caused a startling stun to both society and economy in numerous nations, where different industries suffered unequally. This paper reviews the various performance of the Capital Asset Pricing Model (CAPM), and the Fama-French three-factor model and the five-factor model in different regions and industries. To metric the performance, various statistics models and scaling are applied including Pearson correlation, linear regression, R2 scores, t-test, etc. Specifically, this paper demonstrates the different performances of the CAPM model on the US and Egyptian stock markets, whereas using generalized method of moments in a panel data analysis to evaluate the performance in the U.S. market and the paired sample t-test and Wilcoxon signed-rank to evaluate the performance in the Egyptian market. The Fama-French three-factor model and five-factor model are both based on the U.S. market and analyze the model's performance (measured by significant level) in the U.S. market in general and in individual sectors, respectively. Whereas, in terms of three-factors model, the OLS estimation and relapse expected excess return are used onto the variables and multiple linear regression method was used to study the significance of factors in three sub-industries. Regarding to five-factors model, a multivariate regression with covariates and OLS estimation are the method for evaluation. These results shed light for deeply understanding the model and recognizing the impact on the security market of the COVID-19.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104S (2023) https://doi.org/10.1117/12.2671201
In the field of computer vision, deep learning has developed tremendously, large-scale preforming has received increasing attention from experts and researchers. Different training models often have large performance gaps in training speed and accuracy when performing large-scale pre-training. In this case, choosing the appropriate model for large-scale pre-training is particularly important. This experiment uses the same image data set and the same hardware conditions to construct the image classification model respectively in the three mainstream image recognition large-scale pre-training models, Vision Transformer (VIT), Swin-Transformer and ConvNeXt, try to analyze the advantages and disadvantages of each model by experimental results. It is observed that Vision Transformer has the fastest running speed in computer vision classification experiments, but its accuracy is not as good as the other two models, Swin-Transformer has the slowest speed and average accuracy, ConvNeXt has the highest accuracy, but its speed is mediocre. The results of this experiment have some reference significance for future model selection for large-scale pre-training tasks in computer vision, this can decrease training time and improve training accuracy to some extent.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104T (2023) https://doi.org/10.1117/12.2671333
In this paper, the scatter plots of ethylene selectivity and temperature changes are firstly studied. It can be found that some functional relationships consistent with the scatter trend. By nonlinear regression analysis, the relationship models between ethylene selectivity and temperature are proposed under different catalyst combinations. For the ethanol-coupled preparation of ethylene, the performance of preparation provides a reference for selecting catalysts to design optimal performance reactions. Compared with the original regression model, the final chosen regression function in this paper diverges much lesser from the actual data and the fitting efficiency is better.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104U (2023) https://doi.org/10.1117/12.2671055
This paper constructs a set of product competitiveness evaluation system based on IPVFWA operator. Through literature survey and statistical analysis, the evaluation index system of product competitiveness is established, which includes product development ability index, sustainable value creation ability index, product influence index and loyal customer acquisition ability index; Secondly, IPVFWA operator is proposed to aggregate the evaluation matrix of multiple decision makers into a comprehensive evaluation matrix to determine the evaluation index weight. Finally, a TOPSIS decision-making method based on IVPFWA operator is given, and an example is given to verify the feasibility and effectiveness of the application of IVPFWA operator in product evaluation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104V (2023) https://doi.org/10.1117/12.2671454
We propose a PAMGCN model, which predicts reasonable prescriptions of ancient Chinese medicine. It is used in TCM knowledge graph, according to symptoms described to recommendation a reasonable prescription of traditional Chinese medicine. In the PAMGCN model, we first use multi-GCN to extract patient features, then through the perceptual attention mechanism, to emphasize the influence of patient characteristics on diagnosis and prescription in the model. In the experiment, SMGCN, PMGCN and our model were compared. Our model improved the accuracy by approximately 15.95%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Artificial Intelligence Algorithm and Neural Network
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104W (2023) https://doi.org/10.1117/12.2671526
Under the global pandemic, the application of loans in various forms increases under the influence of economic gloom. Small loan agencies in regions with undeveloped credit systems often neglect background checks because it is costly and time-consuming. Lack of loan approval standards increases the occurrence of defaults. Ideally, a checker program that can make accurate default predictions with a small amount of data contributes to this dilemma. This paper shows a random forest classifier model can achieve 93.2% accuracy incorporating an active learning strategy with only 50 labeled data. Compared to the standard classifier model, the active learning strategy improves the efficiency by 7% with 50 labeled data and increases as much as 14.5% when provided with only 20 labeled data. The source data used in this paper comes from a dataset provided by Univ. AI simulates real-life personal financial records in India.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104X (2023) https://doi.org/10.1117/12.2672682
A more accurate target detection model is proposed in this research based on Yolov5 target detection algorithm, aiming at its low regression accuracy to the target boundary box. Firstly, coordinate attention mechanism is added to the backbone network to improve the position information of the perceived target in the underlying feature information. Secondly, GIOU is replaced with EIOU to improve the convergence speed. Finally, the feature extraction network is replaced with BiFPN to more efficiently fuse different feature information. Using PASCAL VOC 2007 and 2012 datasets and redividing the training set and verification set, this algorithm is better than the original algorithm mAP@0.5 increased by 2.9%, mAP@0.5:0.95 increased by 1.4%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104Y (2023) https://doi.org/10.1117/12.2671738
Idling start-stop is a fuel prudent and emission reductive technology at motor vehicle idling condition. Irrespective of the fuel consumption associated with useless idling condition and the characteristics of actual road conditions, the idling start-stop system not only cannot be fuel-prudence or emission -reduction, but also aggravate the starter abrasion. The method based on BP Neural Network is proposed to predict idling condition in this paper for avoiding useless idle situations. A predictor based on BP Neural Network which has 4 signal-input channels and 1 signal-output channel, is used to predict the speed and idling stop temporal information which is useful in the idling start-stop control policy. The simulation experiment results show that the method based on BP Neural Network can effectively avoid useless idle situations, sequentially reduce fuel consumption and harmful gases discharge and improve comfort.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126104Z (2023) https://doi.org/10.1117/12.2671572
The morphological changes of the hippocampus in brain Magnetic Resonance Imaging (MRI) images are of great significance for the early screening of Alzheimer's disease. Currently, in clinical practice, the diagnosis of the hippocampus is achieved manually by doctors with experience. Because the hippocampus has the characteristics of small size, complex shape, and indistinct boundary with surrounding structures, manual segmentation, and grading of the hippocampus in brain MRI is time-consuming and labor-intensive, which is susceptible to errors because of human subjective judgment. To address that, this paper proposes a hippocampal MRI diagnosis algorithm based on Faster R-CNN and Mask R-CNN. The main contributions are 1) automatic identification of hippocampus in brain MRI by Faster R-CNN neural network, 2) precisely segmenting the hippocampus and judging the atrophy level through Mask R-CNN. Case studies are performed on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database and the medical records of the First Affiliated Hospital of Chongqing Medical University. Results indicate that the proposed method achieves a good segmentation effect on the hippocampus in the coronal MRI image of the brain and accurately grades the level of hippocampal atrophy, which can better assist doctors in diagnosing Alzheimer's disease.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261050 (2023) https://doi.org/10.1117/12.2671072
The particle swarm optimization algorithm (DDPGPSO) is used to study the aiming point optimization problem of multi-bomb attack on complex building targets, using the fire damage model of complex three-dimensional building targets under multi-bomb attack as an example. The theoretical model of the building stereo target is created, the appropriate fire damage index is chosen, and the optimization evaluation function is created, resulting in the optimization model of the optimal aiming point. To address the issue of traditional particle swarm optimization easily falling into local optimization due to premature maturity, a neural network is built to dynamically generate the required particle swarm optimization parameters, reducing the problems caused by manual selection. This algorithm outperforms traditional particle swarm optimization in terms of convergence speed and optimization accuracy, and it has a wide range of applications in the aiming point optimization model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261051 (2023) https://doi.org/10.1117/12.2671145
Federated learning enables multiple parties to jointly train a global model without sharing the original data, which has attracted much attention. Existing research work shows that even sharing local gradients will leak local data. What's worse, the server may deliberately tamper with the aggregation results, resulting in user privacy leakage or other attacks, so users need to verify the correctness of the calculation results returned by the server. In this paper, we design a verifiable privacy-preserving scheme where the server is honest and curious but has the additional ability to forge the aggregated results. The proposed scheme can guarantee the privacy gradient of honest users under the condition that no more than t users collude with the server. During the execution of the protocol, the user is allowed to drop out at any phase, and the aggregated results is kept secret from the server. In addition, each user can verify the correctness of the server’s calculation results, which is the ciphertext of the aggregated results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Xiaomei Hu, Jiahong Weng, Jianfei Chai, Mingnan Zhang, Yilin Li
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261052 (2023) https://doi.org/10.1117/12.2671191
In order to avoid the occlusion problems and missing important features of streamlines in the flow field, this paper proposes a 3D streamline selection algorithm based on comprehensive similarity. The method starts with a hierarchical clustering of streamline sets and then extracts streamlines with high similarity based on their comprehensive similarity. The comprehensive similarity of streamlines requires the calculation of the distance and contour similarity of the streamlines. In this paper, the Hausdorff distance is improved by proposing a partially matched Hausdorff distance to reduce the influence of streamline length on the similarity calculation. Then the contour similarity is calculated according to the ICP algorithm, and the entropy weighting method is used to calculate the weights to obtain the combined similarity. The final comparison with the result of another algorithm shows that the flow field structure is clearer, more complete and more evenly distributed when the streamlines are selected using this algorithm.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261053 (2023) https://doi.org/10.1117/12.2671764
Sleep is the life instinct of human beings. It is not only of great significance to the physical and mental health of individuals, but also can be used as a natural means of regulating, restoring and enhancing bodily functions. There is a prominent contradiction between health needs and the backward status of daily sleep health management, and there is an urgent need to develop theories and methods from sleep structure conversion mechanism to sleep quality monitoring and intervention. In the past few years, artificial intelligence (AI) technology has rapidly emerged in the field of sleep medicine. The purpose of this article is to provide a brief overview of relevant terms, definitions and use cases of artificial intelligence in sleep medicine. AI has a variety of applications in sleep medicine, including sleep and respiratory event scoring in sleep labs, diagnosis and management of sleep disorders, and population health. Although still in its infancy, there are still challenges that hinder the ubiquity and broad clinical application of AI. Overcoming these challenges will help seamlessly integrate AI into sleep medicine and enhance clinical practice. AI is a powerful tool in healthcare that can improve patient care, enhance diagnostic capabilities, and enhance the management of sleep disorders. However, before existing machine learning algorithms can be incorporated into sleep clinics, these artificial intelligence devices need to be regulated and standardized.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261054 (2023) https://doi.org/10.1117/12.2671690
Path planning algorithm is the basis of unmanned ground platform to realize unmanned driving function. Traditional path planning algorithms mostly regard path planning as a geometric problem, which has great limitations on the work of unmanned platforms in the current complex environment. The reinforcement learning algorithm focuses on online planning and has the advantage of continuing to explore and find better solutions on the basis of effective actions. This paper studies path planning of unmanned ground platform based on reinforcement learning method. Aiming at the problems of low flexibility and slow convergence of the current reinforcement learning method in path planning, this paper improves the Q-learning algorithm based on the reinforcement learning algorithm and conducts simulation experiments and analyzes the experimental results. The analysis shows that the path planning algorithm of unmanned ground platform based on reinforcement learning has obvious advantages in performance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261055 (2023) https://doi.org/10.1117/12.2671437
With the continuous progress of new energy technology, new energy vehicles have gradually become the mainstream. More and more new Energy Vehicles (EV) are put into use, which not only brings challenges to the current Smart Grid (SG) system, but also brings brand new opportunities. Vehicle-to-grid (V2G) is a new technology, which not only realizes the power transmission from the grid to the Vehicle, but also realizes the auxiliary power transmission ability from the vehicle to the grid. But the privacy problem and the fairness of the transaction are still an urgent problem to be solved. Privacy issues can lead to the user's privacy being obtained and analyzed by malicious participants, resulting in the user's property being threatened. Fair exchange means that in the transaction process, if one party is malicious, the honest party has no loss or the loss is negligible, or if both parties are malicious or honest, there is no loss for either party. This paper proposes a Vehicle-to-grid privacy protection fair exchange system based on blockchain. Homomorphic commitment and hash chain are introduced to achieve privacy protection and fair exchange.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261056 (2023) https://doi.org/10.1117/12.2671089
Fire is a great danger to the society and economy as well as human life safety, and a small spark can lead to a serious fire. In order to improve the reliability of natural gas station fires warning system, this paper employ YOLO v5 target detection algorithm to investigate the detection of station flames. Firstly, 5488 official datasets and flame images searched for the Internet are collected as training sets and labeled. Secondly, the labeled images are used to train the YOLOv5 network model under Linux operating platform to get the appropriate weighting coefficients to minimize the value of the model loss function. Finally, the trained model is used to identify and detect the flames in the actual natural gas station site environment. The experimental results found that the YOLOv5 algorithm model can achieve real-time supervision of flames, with recognition average precision up to 89.2, fast flame detection and high recognition sensitivity. This work helps to facilitate the realization of real-time monitoring of flames and other hazardous factors of natural gas stations and improve station safety.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261057 (2023) https://doi.org/10.1117/12.2671253
The traditional neural network collaborative filtering algorithm is developed from the matrix decomposition algorithm, and compared with the matrix decomposition, the neural network collaborative filtering algorithm uses a multi-layer perceptron to replace the original dot product operation. Although this method can fully cross user vectors and item vectors, it cannot learn the timing problems on the user side. To solve this, a collaborative filtering model based on deep neural network fusion user sequence is proposed. The model still models the interaction information between the user and the object, with the difference being that recurrent neural networks are introduced when modeling the user. The probability of the user's interest in the item is calculated from the obtained user characteristics and item characteristics. Experiments on the MovieLens and MIND datasets show that the proposed model is higher than the matrix decomposition algorithm and neural network co-filtering algorithm on the AUC and F1 Score indicators, which verifies the accuracy of the model's recommendation effect.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261058 (2023) https://doi.org/10.1117/12.2671059
Relationship extraction is an important task in natural language processing and knowledge mapping. Traditional entity relationship extraction methods have achieved high accuracy in practical applications. However, when faced with the task of entity relationship extraction that is not easy to obtain large-scale supervision training data sets, traditional methods cannot get good results. In this paper, a few-shot relation extraction method based on multi-level feature metric learning is proposed. This method takes the prototype network as the baseline network to generate a class prototype. Firstly, a multi-level feature extraction module is proposed. This module combines the multi-level features of the text with the multi-level attention mechanism, which can fully extract the features of the text. Secondly, a loss function based on label value and negative sample distance is proposed. This algorithm introduces an evaluation mechanism of negative sample distance on the basis of the prototype network, so that the model can adaptively allocate parameters and improve the clustering ability of small samples. Experiments are conducted on FewRel1.0 which is a small sample relational data set. Experiment results show that compared with other models, our model can improve classification accuracy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Pinyue Li, Xiwei Feng, Chi Zhao, Wei Hou, Hao Chen, Yafei Gui, Jiansheng Wu, Chaoqi Wang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261059 (2023) https://doi.org/10.1117/12.2671151
A deep learning multi-classification algorithm is proposed to address the problem that machines cannot accurately understand the semantic differences of the same words in different sentences in the current research on semantic matching question-and-answer classification algorithms. The HanLP word separation algorithm is used to classify the problem first, then the words commonly used in the sentences are replaced according to the correspondence in the dictionary library, and finally the replaced data are imported into the BERT model for encoding and learning. It was verified that the combination of the HanLP word classification algorithm and the BERT model allowed the machine to understand the user's input test to a greater extent and with a higher accuracy rate than other methods. When applied to real-world scenarios, this method will make the user experience more efficient and convenient. With the rapid development of AI, this method will also help in the exploration of natural language.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105A (2023) https://doi.org/10.1117/12.2671321
In order to solve the problem that the quality of feature point matching and the computational efficiency cannot be achieved simultaneously, this paper proposes twin network feature point matching algorithm based on metric learning. Features and feature descriptors of image blocks is extracted through twin networks, and similarity measure loss function is used to complete feature matching in this paper. The results of network training and testing on HPatches dataset show that the algorithm is helpful to improve the accuracy and matching efficiency of feature matching point pairs.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Chang Jian, Gong Yan, Liu Cheng, Yang Muchuan, Liu Jiayu, Jiang He
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105B (2023) https://doi.org/10.1117/12.2671109
In cloud-based edge computing systems, edge server placement is crucial to many applications response time. Generally, edge server placement is always modeled as a multi-objective optimization problem and solved with integer programming algorithms. However, these algorithms are not well scalable, and parameters used in these algorithms depend heavily on experiences. In this paper, we propose an optimized k-mean based edge server placement method. We use k-mean to cluster system sources and application loads and sort the application load to apply the most applicable server to the application. Experimental studies over synthetic data sets validate effectiveness of the method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105C (2023) https://doi.org/10.1117/12.2671218
Aiming at the low matching accuracy of existing local stereo matching algorithms in weak texture areas, a local stereo matching algorithm based on multi-matching cost fusion and guided filtering cost aggregation with adaptive parameters is proposed. First, use the gradient direction to improve the gradient cost, and calculate the matching cost by combining the gradient cost with the Census transform and color cost. Secondly, the cost is aggregated by the guided filtering of adaptive parameters; Finally, the final disparity map is obtained through disparity calculation and multi-step disparity refinement. The improved algorithm is tested on 15 training sets on the Middlebury3 platform, and the average false matching rates of bad4.0 in all areas and non-occluded areas are 19.9% and 13.2%, respectively, which is improved compared with AD-Census and other algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Yingyuan Du, Tao Wu, Gaoyuan Yang, Yuwei Yang, Ge Peng
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105D (2023) https://doi.org/10.1117/12.2671050
Poisoned by the edible fungus accident occurred frequently in recent years since that there were no effective and quick recognition methods for the wild fungus. To tackle the problem, a wild fungus classification algorithm based on a deep convolutional neural network (CNN) and Residual Network (ResNet), is proposed in this paper. An optimization method is also proposed for network training. In order to verify the effectiveness of the model and optimization method, a wild fungus database, in total of 1280 images, is used in this paper. The experimental results show that the proposed algorithm can effectively complete the classification task of wild mushrooms, and the optimization algorithm proposed in this paper can also effectively improve the classification effect of the algorithm model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105E (2023) https://doi.org/10.1117/12.2671090
Aiming at the problem of unclear human-object interaction behavior objects in complex background, we propose an Instance-Based Adaptive Attention (IBAA) algorithm. The algorithm adaptively generates a series of attention matrices according to the objects and subjects of human-object interaction instances. These attention matrices are then used to update the feature map to reduce the interference of the complex background. In order to enrich the human object interaction information, we use the graphical model to represent the interactions between human and objects and use the graph convolutional neural network to update it. Experimental results on HICO-DET dataset show that the proposed algorithm has significantly improved accuracy and multi-scale object detection ability compared with other human object interaction algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105F (2023) https://doi.org/10.1117/12.2671202
To address the current problems of wide range of strip steel surface defect size variation, slow detection efficiency, low detection accuracy, and difficulty of mobile-side model deployment, an improved YOLOv4 algorithm model is proposed in this paper. Firstly, in order to improve the robustness of the model, data augmentation is applied to the dataset. Secondly, in order to improve the matching between the a priori frame and the feature map, the K-means++ algorithm with faster convergence and better results is used instead of the K-means algorithm in the original YOLO algorithm for the design of the a priori frame. Finally, CSPDarknet is specifically replaced for the Ghostnet to enhance the backbone network's ability to extract defective features. The experimental results show that the improved YOLOv4 algorithm achieves 87.9% mAP on the publicly available NEU-DET dataset, which is 2.4% lower than the original YOLOv4 algorithm. However, the number of parameters of the model decreases by 80% compared with the original YOLOv4, and the detection speed is around 44 FPS, which can not only meet the needs of industrial production, but also meet the requirements of deploying the model to mobile.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105G (2023) https://doi.org/10.1117/12.2671320
The traditional object detection network (such as r-cnn, faster r-cnn) has low detection efficiency and accuracy, which leads to the low speed, low accuracy and excessive target re-recognition times of deep sort object tracking algorithm. In this paper, the yolov5 network is selected as the object detection framework of deep sort object tracking algorithm, and the yolov5 algorithm is improved with the temporal difference. Meanwhile, the distance matching mechanism and feature extraction network of deep sort algorithm are improved. Experimental results show that the improved method can improve the accuracy of object tracking algorithm and reduce the number of target re-recognition.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Peng Hang, Yan Yan, Xianlan Fu, Haishan Chen, Yingjie Liu
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105H (2023) https://doi.org/10.1117/12.2671069
In order to make the intelligent vehicle run safely and improve the applicability of local path planning algorithm in intelligent vehicle, the paper presents a local path planning method for intelligent vehicles. Firstly, establish vehicle model of Ackerman steering. In addition, the minimum turning radius constraint is added to the speed screening mechanism based on the traditional dynamic window approach. Then, in order to avoid excessive changes in the driving speed of intelligent vehicles, the curvature retention evaluation index is added to the evaluation function. The simulation results show that the improved algorithm meets the requirements of intelligent vehicle dynamic obstacle avoidance and can plan a safe and reasonable path.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105I (2023) https://doi.org/10.1117/12.2671397
Faulty recognition and recovery are extremely essential components for existing sensor networks. However, current investigations concentrate on the recognition process and develop abundant advanced identification methods and existing recovery method is almost based on the multiple layers of neural network to achieve the recovery of sensor network, which would result in vast computation costs. Attention mechanism is proposed for neural network to focus some important neural nodes and perform high accuracy for training process with insignificant computation cost. In this article, we propose a novel neural network to recover the fault sensor network with attention mechanism. Thus, the developed mechanism can achieve rapid recovery of fault sensors network with reasonable computation costs and acceptable performance. From extensive experimental result, we can conclude that our proposed method can strictly process the fault sensor network and compared with existing fault sensor network recovery methods to illustrate the effective performance on recovery accuracy with reasonable costs of our proposed method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105J (2023) https://doi.org/10.1117/12.2671176
At present, the academic community has carried out some research on knowledge reasoning using Reinforcement Learning (RL), which has achieved good results in multi-hop reasoning. However, these methods often need to manually design the reward function to adapt to a specific dataset. For different datasets, the reward function in RL-based methods needs to be manually adjusted to obtain good performance. To solve this problem, an agent training model combined with Generative Adversarial Networks (GAN) is proposed. The model consists of two modules: a generative adversarial inference engine and a sampler. The sampler uses a policy-based bidirectional breadth-first search method to find the demonstration path, and the agent uses the reward considering the information of the neighborhood entities as the initial reward function. After sufficient adversarial training between the agent and the discriminator, the policy-based agent can find evidence paths that match the demonstration distribution and synthesize these evidence paths to make predictions. Experiments show that the model achieves better results in both fact prediction and link prediction tasks.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105K (2023) https://doi.org/10.1117/12.2671056
In optical cameras, there are deviations in the manufacturing process and assembly accuracy of image sensors, lenses and other related components, which will produce various forms of image distortions, seriously affecting the application and development of Computer Vision (CV) in real life. In order to solve these problems, a hybrid solution algorithm (GPSO-ILM) based on Global Particle Swarm Optimization (GPSO) and improved Levenberg-Marquardt (LM) is proposed on the basis of Homography Transformation, which transforms the solution of nonlinear distortion equations into a parameter optimization problem and can quickly iterate to obtain the optimal solution of the target function. It can effectively avoid the problems that traditional LM depends on initial values and is easy to fall into local convergence. In order to verify the effectiveness of the algorithm, the improved GPSO-ILM method is compared with the Tasi method of two-steps and traditional Zhang's correction method in OpenCV. The experimental results show that the RMSE errors of the improved GPSO-ILM method are relatively reduced by more than 23% and 18%, and the processing time is saved by about 25%. Compared with other methods, this method has higher correction accuracy and faster processing efficiency.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105L (2023) https://doi.org/10.1117/12.2671712
A multifunctional intelligent crutch is designed for the travel safety problem of the elderly group. The STM32F103C8T6 microcontroller is used as the main controller, it can realize obstacle detection, fall detection and step counting functions combined with ultrasonic sensors and acceleration sensors, fall detection and step counting functions, SIM808 is used to locate the user's position and connect to the cloud server, when user falls, the fall state and GPS location information will be sent to cloud based on the MQTT protocol. The test results show that the system is safe, reliable, simple and convenient, it is appropriate for monitoring the travel conditions of the elderly group, and it can greatly improve the travel safety of the elderly group.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Qingbo Ji, Yufei Qi, Hang Liu, Changbo Hou, Jianyang Gong
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105M (2023) https://doi.org/10.1117/12.2671078
In the field of computer vision, object tracking is a hot research topic. At present, the coordinate box of most algorithms is vertical box. However, in practical application, the frame selection area of the object is expanded due to the rotation of the object, which makes it impossible to obtain the real motion angle of the object. To solve this problem, this paper proposes a Siamese network tracking algorithm (SiamRotate) based on rotation box reinforcement learning. In order to predict the rotation Angle of the object, firstly, we process the VOT2013 and VOT2014 datasets to increase the angle richness and used the processed datasets to train the angle decision-making network through the combination of supervised training and reinforcement learning. Based on the SiamRPN network, the tracking network obtains the feature response through the deep cross-correlation operation of the feature map of the two branches through SiamRPN, and then processes the feature response map through the angle decision network to predict the object rotation angle. The performance indexes of the proposed method reached 0.62(A), 0.21(R), 0.425(EAO) on VOT2016, and 0.61(A), 0.27(R), 0.399(EAO) on VOT2018, and the tracking speed reaches 138fps. Finally, comparative experiments show that the algorithm has good performance in speed and accuracy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Juan Fang, Qiangang Zheng, Weimin Liu, Haibo Zhang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105N (2023) https://doi.org/10.1117/12.2671152
With the development of Reinforcement Learning (RL), it becomes able to solve the continuous action space problem and shows strong ability in dealing with complex nonlinear control problem. Based on the Deep Deterministic Policy Gradient (DDPG) algorithm, a novel scheme of aeroengine acceleration controller is proposed in this paper. According to the characteristics of the engine acceleration stage, the reward function is constructed, and the state parameters are updated in the form of sliding window to reduce the sensitivity of the network to noise. DDPG adopts actor-critic framework, critic calculates value function by the deep neural network, actor outputs action command and forms a closed-loop control system with the engine. The method is verified by digital simulation at ground condition and the results demonstrate that compared with the traditional PID controller, the acceleration time of DDPG controller is reduced by 41.56%. Additionally, the network converges within 400 steps.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105O (2023) https://doi.org/10.1117/12.2671412
Nowadays, deep learning models are widely used in medical diagnosis, with advantages of higher accuracy and better capability of analyzing complex medical images over traditional check. This work compares seven commonly used CNN models - VGG16, ResNet50, InceptionV3, DenseNet-121, Xception, Inception-Resnet V2, and MobilenetV3-small - on the diagnosis of the four early stages of Alzheimer’s Disease – Normal Control (NC), Early Mild Cognitive Impairment (EMCI), Mild Cognitive Impairment (MCI), and Alzheimer’s Disease (AD). Models are initially adjusted with a dense layer of four outputs on the multi-classification problem and deployed in various overfitting prevention setting. Also, by using transfer learning and fine-tuning technique, the training accuracy and testing accuracy of the seven models have both improved by over 30%. Among all, DenseNet121 has the overall best performance of 95.48% testing accuracy, 0.92 F1- score and second smallest number of parameters of 7,041,604, with a good portability on mobile computing devices, likely due to the reuse of feature maps to extract more accurate features. Inception-ResNet-v2 is adopted when only considering accuracy and reliability and due to its largest number of parameters, Inception-ResNet-v2 is more suggested to be used on computer medical application. VGG16 has the highest accuracy of 96.11% and a small number of parameters of 14,716,740, however, it is currently not suitable for applying to real-life application since its F1-score of 0.8943 and precision of 0.8266, indicating numerous false cases.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105P (2023) https://doi.org/10.1117/12.2671082
Aiming at the problems of personnel intrusion and random placement of goods affecting driving safety in the process of forklift operation, an obstacle target detection and distance detection method for forklift operation based on improved YOLOv5 network is proposed. First, in view of the problem that the target is small and difficult to detect when the distance between people is long, the YOLOv5 network is improved to improve the accuracy of target detection; for the problem of tracking loss caused by the detection target being occluded, the target tracking algorithm and ranging algorithm are introduced to realize the target detection tracking recognition and distance detection. The experimental results show that the detection accuracy of the improved YOLOv5 network is improved by 4%; within the dangerous distance range, the distance detection error is within 5%, which meets the requirements of real-time detection and distance detection and can be used to detect potentially dangerous situations during operation and give an early warning.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105Q (2023) https://doi.org/10.1117/12.2671172
In order to improve the effect of ring network cabinet life prediction, the deep neural network life prediction method based on the characteristics of internal distribution of ring network cabinet is studied. Using the optimal wavelet packet transform method, the local discharge characteristics of ring network cabinet are extracted. Nuclear principal component analysis was used to reduce dimension to deal with the local discharge characteristics of ring network cabinet. The bidirectional long-term memory deep neural network was established. The local distribution characteristics after dimensionality reduction were input into the network and the autoregressive comprehensive moving average model, and the life prediction results of the ring network cabinet with nonlinear and linear characteristics were output. The final life estimation results are obtained by combining the two estimation results. Experimental results show that the algorithm can effectively extract and reduce the dimension of the internal local discharge features of ring network cabinet. It can accurately predict the service life of the ring network cabinet under different types of local distribution. Under different local discharge intensities, the R-square coefficient of the algorithm for predicting the life of the ring network cabinet is higher, which has better prediction effect.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105R (2023) https://doi.org/10.1117/12.2671075
Aiming at the problems of blindness, slow convergence and easy to fall into local optimal solution in the early stage of traditional ant colony algorithm search, an improved ant colony algorithm was proposed. The algorithm first introduces the gravitational potential field function, constructs new heuristic information, and improves the convergence speed. Secondly, learn from the wolf pack update rules, improve the pheromone update function, and give full play to the feedback function of pheromone. Finally, the path is optimized locally to remove redundant nodes. The comparison experiment with the traditional ant colony algorithm proves that the improved algorithm shortens the running time, the convergence speed and the global search ability are better than the traditional ant colony algorithm, and it is not easy to fall into the local optimal solution, which verifies the effectiveness of the algorithm.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Chenhuan Tang, Shiran Zhu, Meng Zhang, Jie Chen, Xingyi Guo
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105S (2023) https://doi.org/10.1117/12.2671703
Based on YOLOv4-tiny, A lightweight mask detection algorithm is presented. By replacing the CBL module in the backbone feature extraction network (CSPdarknet-tiny) and Yolo Head with Ghost module that reduces the parameters of the network model. By the combination of Ghost module, CBAM attention, SMU activation function, and BN layer, a lightweight attention mechanism residual module (GCS_Block) is designed, which is embedded into the backbone feature extraction network, improving the model extract mask feature level. The Kmeans++ method is used to perform anchor box clustering on the dataset in this thesis. The experimental results show that compared with YOLOv4-tiny, the MAP has increased from 74.02% to 86.77%, the parameter has decreased from 6,056,606 to 1,657,828. The memory size of the model is 5.6MB.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105T (2023) https://doi.org/10.1117/12.2671249
China is the world's largest emerging energy market, but due to the rapid development of China's new energy industry special, the value of new energy companies are more susceptible to sharp increases and decreases brought about by external factors in the market compared to other companies. In this paper, we propose an algorithm that uses an effective cluster learning strategy (Random Forest) contrasted with a genetic algorithm. A new financial forecasting model GSRF is constructed by parameter optimization of random forest model through grid search algorithm, and the model is applied to short-term stock forecasting. This paper builds a stock price trend forecasting model based on both algorithms and uses the model to determine whether the cost of a stock will be higher than its cost at a given date. The experimental results show that the model built using GSRF stochastic forest has the highest return and the lowest risk compared to the return based on the traditional genetic algorithm trading strategy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105U (2023) https://doi.org/10.1117/12.2671477
This study examines the literature on AI and libraries, examines the significant roles AI has played recently in industries related to libraries, and briefly describes relevant technical functions and their application characteristics in the library field. It begins with six key technologies: OCR, data mining, natural language processing, face recognition, knowledge mapping, and machine learning, and then makes a thorough analysis of each. Detailed analysis and summary of the results achieved in the practical application of AI, an analytical overview of the business functions related to AI in the library field on the development and reform of libraries and the current application status of various technologies, and the problems that libraries may encounter in the practical implementation of AI-related technologies are pointed out.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105V (2023) https://doi.org/10.1117/12.2671395
When it is used to cluster datasets with complex structure, the Affinity Propagation (AP) algorithm faces a number of problems such as excessive local clustering, low accuracy, and invalid clustering evaluation results of some internal evaluation indexes due to excessive clustering. In view of this, this paper proposes an algorithm designed to determine the optimal clustering number. In this paper, the methods of coarse clustering and merging similar clusters are adopted to reduce the clustering number and optimize the maximum clustering number (Kmax), and new calculation methods for intra-cluster compact density, inter-cluster relative density and cluster separation are provided, based on which a new internal evaluation index is designed. The experimental results regarding UCI and NSL-KDD datasets show that the proposed model can provide correct clustering partitioning and accurate clustering range and can well outperform the other three improved algorithms in relevant detection indexes such as detection rate and false alarm rate.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105W (2023) https://doi.org/10.1117/12.2672167
Isonicotinic acid is used as a pharmaceutical intermediate, mainly for the production of the anti-tuberculosis drug isoniazid. Prediction of isonicotinic acid yield using data from the production process is helpful to ensure product quality and improve production efficiency. Traditional BP neural networks have lots of disadvantages such as slow convergence, easy to fall into local minima and sensitive to the selection of initial weights and thresholds. In order to predict isonicotinic acid yield efficiently and accurately, a prediction model of isonicotinic acid yield based on the Grey Wolf Optimizer (GWO) optimized BP (GWO-BP) neural network was proposed. The prediction model was used to predict the historical production data of isonicotinic acid in a plant, and the experimental results showed that the accuracy of the proposed GWO-BP prediction model was higher compared with the traditional BP and GA-BP prediction models.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105X (2023) https://doi.org/10.1117/12.2671166
In this paper, the aerodynamic loss coefficient of the airfoil is obtained through GA-BP Neural Network. The maximum thickness, the position of the maximum thickness, the blade camber angle and the incident angle are set as input parameters, and the loss coefficient is output parameter. The neural network optimized by genetic algorithm is used for training and testing. The GA genetic algorithm is used to optimize the operating conditions, and the data is imported into the prediction model established by the BP neural network for training, and an effective loss coefficient prediction scheme is obtained. The research results show that the GA-BP neural network has high prediction accuracy, and the mean square error of prediction is 6.4228e -05, which can effectively solve the loss coefficient prediction problem of airfoil.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Haoqing Li, Xiaohao Huang, Changchun Pan, Chunlei Yang, Jinbao Wang
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105Y (2023) https://doi.org/10.1117/12.2671282
As a key indicator in ship design, many major incidents of ship sinking are related to the ship's damaged stability. The process of calculating the damaged stability becomes more and more complex and time-consuming on account of more and more stringent specification standards. A two-stage design step is used in this article to realize the calculation of ship’s damaged stability under various watertight bulkhead fast. Firstly, a multi-layer feed-forward neural network model was designed for the predictive regression of a ship's damaged stability using the location of the watertight bulkhead as a variable. Secondly, the relationship between each watertight bulkhead variant and the damaged stability A-value is analyzed. After that, with hydrostatic curve calculation based on the inlet simulation and the interaction between watertight bulkheads considered, a multilayer feed-forward neural network model based on the attention mechanism is designed, which could predict the regression of the damaged stability A-value and analyze bulkhead weights. Finally, the validity of the model was verified by the data, in which the mean value of the prediction error MAE (mean absolute error) was at 2.67×10-4 and the computation time was greatly reduced.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105Z (2023) https://doi.org/10.1117/12.2671216
In recent years, stock price prediction has become a research hotspot. The price of the stock market is unstable, which often rises or falls sharply due to the national policies, which makes it difficult for investors to achieve stable returns in the stock market. With the rapid rise of artificial intelligence, computers have become flexible in dealing with mathematical problems. Therefore, the extraordinary computing power of computers has been used to analyze and predict the trend of the stock market. More and more computer professionals began to enter the financial market and use neural network to study the trend of the stock market. This paper uses BP neural network and LSTM neural network to learn and predict the stock data of Shanghai Composite Index from January 2012 to June 2022. LSTM is a kind of RNN, but it is superior to other neural networks. It can effectively deal with data forgetting and gradient explosion problems and bring reliability to the prediction results of the model. The two models are evaluated by analyzing MAE, MSE and the time required for model training. The results show that LSTM model can not only learn longer time span than BP model, but also better than BP model in MAE and MSE indexes, which provides some reference and guidance for the prediction of medium and long-term stocks.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261060 (2023) https://doi.org/10.1117/12.2671670
It is not uncommon for bridges to be damaged by earthquakes. As an important throat in the transportation network, earthquakes not only cause the loss of bridges themselves, but also cause a series of losses of the transportation network. In the design of bridge seismic isolation, the seismic isolation device is used to isolate the structure from the frequency band where the seismic energy is concentrated and reduce the seismic response of the structure by prolonging the period and increasing the damping. Compared with advanced countries in seismic research, there is a big gap in the research of mechanical model and parameters of seismic isolation devices in China's bridge seismic code. Based on the analysis of the advantages and disadvantages of the finite element sensitivity method and its application limitations, a sequential recurrence method for support optimization is proposed. The results of an example show that the sequential recurrence method has the advantages of strong adaptability and unconditional convergence. On the basis of the sequence recurrence method, the modified sequence recurrence method is further proposed. This method can take into account the influence of the pier inertia force and is applicable to a variety of support forms. After the genetic algorithm optimization calculation, the difference of the bending moment at the bottom of each pier can be controlled within, which solves the problem of large difference of the bending moment at the bottom of the pier under the longitudinal earthquake of irregular bridges.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Tang Yu, He Jing, Chen Shijun, Li Yongfang, Wang Shaoxiong, Fang Yue
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261061 (2023) https://doi.org/10.1117/12.2671039
In this thesis, the path planning for the pirate attack incidents encountered during the navigation of the ship is firstly carried out. On this basis, the path planning of other dynamic obstacles in the pirate attack area is further considered. On this basis, the path planning of other dynamic obstacles in the pirate attack area is further considered. The grid method is used to model the research sea area, and the dual-objective model is constructed to ensure the minimum fuel cost and risk cost at the same time. The grid method is used to model the research sea area, and the dual-objective particle swarm algorithm to solve the problem. calculation results based on the optimal solution of pareto are better than the results obtained by the weight method, and the results are simulated on a It is concluded that the calculation results based on the optimal solution of pareto are better than the results obtained by the weight method, and the results are simulated on a large-scale navigation simulator, which proves the validity and feasibility of the model. The results are simulated on a large-scale navigation simulator, which proves the validity and feasibility of the model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261062 (2023) https://doi.org/10.1117/12.2671046
Traditional forecasting models mainly utilize historical sales and a few macroeconomic indicators, can no longer reflect the impact of new energy vehicle sales. Based on high-dimensional tensor CNN, this paper proposes the new-energy vehicle sales prediction model and explores the prediction effect of high-dimensional data and deep learning on new energy vehicle sales, the factors that affect sales is divided into consumer dimension, vehicle dimension and social dimension. In each dimension, we choose 25 high comprehensive influence factors, then integrate them into a tensor structure. Through the one-dimensional multi-kernel convolutional neural network, the correlation between sales and tensor data is learnt. Furthermore, thirty vehicle brands with sales advantages in recent three years are verified. The experimental results show that, compared with linear regression, random forest, light gradient boosting machine and convolutional neural network, that commonly used in sales forecasting.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Proceedings Volume Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 1261063 (2023) https://doi.org/10.1117/12.2671174
To address the problem of time delays in the information dissemination process, the classical Hegselmann-Krause model is extended by introducing concepts such as time delays. The role of network models, trust thresholds and time delays in the opinion evolution process is explored. The results show that scale-free networks are more likely to reach opinion consensus than random networks under the same trust threshold, the larger the trust threshold in the same network model the easier it is for opinions to reach consensus, and conversely the smaller the trust threshold the more fragmented the opinions tend to be. secondly, time delay speeds up the initial stage of opinion evolution. Although this feature is counterintuitive, the value of this feature is verified in example simulations. The simulation results considering time delays are closer to the evolutionary process of the real case.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.