This PDF file contains the front matter associated with SPIE Proceedings Volume 12474 including the Title Page, Copyright information, Table of Contents, and Conference Committee Page.

SM2 digital signature algorithm (SM2-DSA) is the Chinese version of the elliptic curve digital signature algorithm (ECDSA), which has become one of the international standards of elliptic curve cryptography. Despite its solid theoretical security, SM2-DSA is still prone to a variety of physical attacks. Hence, it is important to research the security of the SM2- DSA implementation. In this paper, we propose a fault attack model for the SM2-DSA based on the weak elliptic curve. Experimental results show that the proposed model can directly calculate the parameters of the fault curve by using the fault signature pair, and if the fault injection location is correct, we only need an error signature pair to recover the 256-bit signature private key within 3 minutes. Compared with the general weak elliptic curve attack, our model is more practical and 20% more efficient in recovering the private key.

Deep learning techniques have been widely used in the field of Side Channel Attack (SCA), which poses a serious threat to the security of cryptographic algorithms. However, deep learning-based side channel attack also has problems such as inefficient models, poor robustness, and longtime consumption. To address these problems, this paper focuses on the performance of Long Short-term Memory(LSTM) combining with the dimensional compression technique of Sparse Auto Encoder (SAE), and validates it on fully synchronized and unsynchronized EM traces captured under first-order bool mask protection. The experimental results show that compared with multilayer perceptron (MLP) and convolutional neural network (CNN), LSTM achieves more than 90% training accuracy and test accuracy, with higher robustness, lower parameters and faster convergence speed, even when the jitter in the dataset increases from 0 to 50 and 100.

At present, with the deep integration of advanced information and communication technologies such as artificial intelligence, big data, and the Internet of Things and power technology, compared with the traditional power grid, the new power system is more multi-source heterogeneous, open and ubiquitous, and its attack surface is greatly expanded, the difficulty of security protection has increased sharply, and it is urgent to improve the security construction of the power information system. Compliance inspection and security assessment are important ways to guide the security construction of power information systems to this end, this paper is in-depth understanding of the current status of the network security of the power information system, combined with the network security classified protection system 2.0. A comprehensive evaluation method of power information system security protection based on entropy weight-TOPSIS algorithm is proposed, and through model case analysis, it provides an important theoretical basis for the safe, reliable and standardized development of power system.

With the continuous development of digital multimedia technology, digital multimedia technology has realized the integration with medical treatment, education, traditional scientific research and other fields, and become the trend of future development. The application of digital multimedia technology to reform vocal music performance teaching is an important means of innovative vocal music teaching, but also an inevitable choice to promote the development of vocal music teaching. Digital multimedia technology will be one of the irreplaceable important carriers in vocal music performance teaching. In the application of digital media, many new teaching modes have appeared in the design of vocal music learning system in colleges and universities. Digital multimedia technology has effectively broken the limitations of traditional vocal music teaching and provided new possibilities for the informatization and modernization of vocal music performance teaching and learning. This paper mainly describes the design of multimedia vocal music learning system based on Visual C++, according to the current problems in multimedia vocal music teaching, and put forward their own solutions, the purpose is to speed up the design process of multimedia vocal music learning system, improve the progress of students learning.

Detecting anomalies in HTTP request data is a vital security task. With big data becoming ubiquitous, techniques for structured graph data have been focused on recent years. As nodes in graphs have long-distance correlations, detecting anomaly in plain structured graph data is practical. This paper proposes a node-level feature-based regression detection method. Given a graph generated from a snapshot of HTTP request data collected by API gateway and considering clustering coefficient and empirical inspired rules, construct a regression model to dig out substantially deviate nodes. Extensive experimental studies on a real-world request dataset demonstrate that it performs relatively prominent and favorably to HBOS (a concurrent density-based method) and iForest (a linear time complexity model-based method with a low memory requirement) in terms of ROC-AUC and processing time.

Software vulnerabilities are an important resource for cyberspace security. The rapid development of automated bug finding methods represented by fuzzing enables vulnerabilities to be found quickly, but the precise analysis of vulnerabilities mainly relies on manual labor. To improve the efficiency of vulnerability analysis, many automated vulnerability analysis tools have emerged in recent years, and how to evaluate these analysis engines has become a new challenge. This paper designs and implements a set of anomaly sample datasets for vulnerability analysis and introduces the construction method of the datasets. The data set has the characteristics of complete variety, strong applicability, and high degree of expansion, and is expected to support the ability verification of vulnerability analysis tools.

With the continuous development of new media and network information technology, more and more reconfigurable resources of network digital media are needed to build a reconfigurable resource allocation model in network space. The dynamic allocation method of network digital media reconfigurable resources based on multi-source feature fusion cluster analysis is feasible to a certain extent. Combined with the result of feature extraction, the dynamic allocation of reconfigurable resources is realized. Research on the reconfigurable resource allocation method of network digital media is of great significance in improving the utilization capacity of network resources. Traditional method, the dynamic network reconfigurable digital media resources allocation methods mainly include dynamic allocation method based on the characteristics of correlation analysis, PCA principal component analysis method of dynamic allocation and dynamic allocation of fusion k-means clustering method and so on, USES the statistical features extraction and autocorrelation detection, realize reconfigurable dynamic allocation of resources, However, the traditional method of resource allocation has poor fitness and weak feature identification ability, so it needs to be changed. This paper mainly describes the dynamic allocation method of reconfigurable resources of network digital media, aiming at strengthening the allocation of network digital media resources, so as to further improve the dynamic allocation ability of reconfigurable resources and speed up the application of network digital media.

This paper analyzes the similarity between steganography and adversarial attack, and then analyzes the attack and defense stage of adversarial samples from the perspective of steganography, which provides a broader idea for researchers in this field. This paper validates the efficiency of steganography on help resisting adversarial attack and proposes a steganalysis based method to enhance the adversarial attack against steganalysis detectors with the help of a distortion function. We combine the distortion map with gradient information to select the disturbed pixels in each attack iteration to generate adversarial samples that can bypass steganalysis detectors. The results show that steganography technology can help detecting normal adversarial samples, and our enhanced adversarial samples achieve a much higher transparency rate against our steganography based method than other normal adversarial samples.

In order to meet the diversified needs of Japanese learning, this study, combined with the concept of Internet +, conducts research on the construction of WeChat public platform for Japanese learning. Firstly, the relevant research concepts and the basic architecture of the platform are determined, and then the database and relevant important technologies applied in the platform are discussed in detail. Finally, on this basis, the expected functional modules of the WeChat public platform for Japanese learning are described, so as to provide reference for future teaching and learning based on the WeChat public platform.

SQL injection attack could obtain sensitive information in the database, tamper or delete illegally obtained information, etc., which causes immeasurable losses to the system. Aiming at SQL injection attack, this paper proposes a new SQL injection detection scheme that combines traditional detection methods with abstract syntax tree structure judgment based on semantic analysis. The solution includes modules such as data preprocessing, SQL statement pre-assembly, and semantic analysis. By assembling the user input content and the actual SQL template statement to form a complete SQL statement, the statement is subjected to structural judgment and semantic analysis to determine the request and precisely identify malicious injection attack.

Aiming at the problems of low consensus efficiency and low scalability in traditional Byzantine consensus algorithm, propose a Practical Byzantine Algorithm Based on Node Reputation Value Matching TMBFT (Trust-based on Node Reputation Value Matching). Select some consensus nodes for consensus voting according to the neighbor matching model. At the same time, a reputation value evaluation mechanism is introduced to supervise the consensus behavior of neighbor nodes. To sum up, compared with the traditional practical Byzantine fault-tolerant algorithm, the consensus algorithm based on node reputation value matching improves the consensus efficiency of the system, reduces the node overhead in the network, and enhances the stability and security of the system.

Traditional 3D printing usually uses professional software such as SolidWorks and Maya for modeling. However, applying the software mentioned above to customize 3D food printing patterns has the problems of a solid professional operation and high skill requirements. In response to this problem, a pattern drawing software applied to 3D food printing equipment was designed. The software uses image processing technology and C# programming language, based on Windows operating system and. NET Framework, combined with SQLite database technology to achieve product pattern drawing and model export functions. Among them, the automatic extraction of the enclosed area realized by the flood filling algorithm is the core function of product pattern drawing; the automatic acquisition of the contour of the connected graph recognized by the canny edge detection algorithm is the crucial step of the model derivation. The test results show that the software can quickly implement template import, pattern drawing, model export, and other operations, which improves the efficiency and ease of operation of pattern drawing in 3D food printing; The model file exported by the software can be directly used in the layered printing process of a 3D printer, which provides a direction for promoting the application of 3D printing technology in the food industry.

FPGA+DSP architecture is widely used in satellite digital processing systems. The satellite is in an environment where there is a lot of particle radiation and collisions. In order to avoid changes in stored code data brought by SEU (Single Event Upset), improve the system reliability, at the same time meet the current requirements of software on-orbit reconfiguration, the FPGA and DSP in the space-borne core processor need to have the ability of error correction and reconfiguration. To do this, FPGA and DSP programs need to be stored in external, writable memory. Nor Flash, with its high capacity and reliability, is often chosen as the memory for storing code. At present, the on-orbit maintenance of FPGA is usually realized by using Actel FPGA to conduct TMR (Triple Modular Redundancy), refresh and other processing on the FPGA code stored in Flash. Based on this idea, a T-shaped structure is constructed among FPGA, DSP and Flash for the architecture of the space-borne processor. As the master, FPGA controls Flash to complete TMR, error correction and on-orbit reconfiguration of DSP code. This method reduces hardware redundancy, gives consideration to autonomous maintenance and on-orbit reconfiguration, and increases system robustness. This method has been applied and fully verified in orbit.

Solar energy is more and more widely used, but photovoltaic is also easily affected by environmental factors. Because of the remote location of the photovoltaic power station, it is very difficult to clean the surface cover manually. However, the use of robot cleaning depends on the accurate identification of the shielding objects. Due to the wide variety of shielding objects, the current technology is difficult to identify accurately. In view of the fact that the types of fallen leaves of photovoltaic panels are complex and difficult to clean, An occlusion detection algorithm for small targets on the surface of photovoltaic modules based on deep learning is proposed, and the model network method for quickly detecting leaf occlusion and determining the occlusion position of photovoltaic panels is discussed. In this paper, an improved YOLO-PX algorithm is proposed to identify and classify the occlusion of photovoltaic modules. Target detection experiments are carried out on the field data set of photovoltaic power station by using the original YOLO algorithm and the improved YOLO-PX algorithm. The experimental results show that the effect of the improved algorithm is good, and the detection accuracy and recall rate are 96.3% and 94.2% respectively. This method can provide technical support for the intelligent operation and maintenance of photovoltaic power station.

On the basis of the standard particle swarm optimization and the bionic theory, the standard particle swarm optimization algorithm is improved based on the biological symbiosis mechanism, and the weight selection strategy and search space of the standard particle swarm are improved. The weighted particle swarm optimization algorithm (Particle swarm optimization based on biological symbiosis mechanism and self-adaptive inertia weight, PSO-BSMSIW) has been verified by the test function, and the algorithm has significantly improved the convergence speed and accuracy.

Under the influence of the epidemic, the offline cooperation between college laboratories has decreased, more and more cooperation is going online. Therefore, the research on the integration and sharing of heterogeneously described university resources has become a hot topic. Based on some scientific metadata standard research, this paper proposes metadata standards including university talents, instruments, laboratories by analyzing the metadata standards of existing scientific data. Then, a heterogeneous scientific data integration mapping algorithm is designed.

Data mining plays an important role in getting insights from news text. This study collected 695,051 English-language news reports on terrorism and counter-terrorism from March 2017 to March 2018 in BBC news and conducted a text analysis with LDA topic modeling. 20 topics and five themes were classified, and it was disclosed that major themes include that: (1) BBC focused on constructing local discourse structure, (2) Comparing the news reports at home and abroad, and (3) Historical origins, the development in ancient and modern times and a trial of strength between different countries.

Aiming at the problem of low analytical efficiency in the current analysis methods of data file format of scientific instruments, a data file format analysis method based on clustering was proposed to improve the efficiency of file format analysis. According to the file storage structure and the characteristics of cluster distribution, the selection principle of file samples in cluster analysis is formulated. At the same time, the corresponding format analysis auxiliary tool software is developed, which can automatically judge the rationality of the selected files and automatically group them, simplifying the corresponding format analysis process. The method and the developed tool are used to analyze the format of MS data generated by a mass spectrometry model. The experimental results show that the format of MS data obtained by this method is accurate and the efficiency is significantly improved. This method can effectively promote the sharing of data resources of large-scale scientific instruments and improve the utilization rate of data resources.

In recent years, with the rapid development of technology fields such as big data, cloud computing, Internet of Things, and mobile Internet, security incidents such as network attacks and data information leakage have occurred frequently, which shows that the current information system falls in the serious security situation, and methods relying on the traditional security protection mechanism to ensure information security has gradually become inadequate. Compared with other software languages, Java language is widely used in the development of large-scale business systems due to its high access, concurrency, and clustering. Source code is the basic element of building a business application system, and logic vulnerabilities or nonstandard programming in code are the roots of application security events. This paper proposes a source code security defect assessment method based on the entropy weight method by deeply analyzing the Java source code security defect detection and repair methods.

Feature extraction is one of the critical technologies in trajectory data clustering. Extracting useful features for trajectory clustering is a difficult problem when there are missing segments in trajectories. This paper proposes a feature extraction algorithm for missing trajectories clustering. The algorithm converts trajectories into images, a multi-layer image preprocessing method is proposed to reduce the information loss during image processing. Then build an autoencoder to extract image features. The loss function is designed according to the attention mechanism to highlight the effective information in the trajectory image. The autoencoder’s ability to handle missing trajectory segments is trained by adding artificial image masking. The effect of feature extraction is verified by clustering. Compared with the unimproved autoencoding and interpolation methods, the clustering effect of the features extracted by this algorithm is improved. At missing rate 50%, this is an increase of eight and six percentage points, respectively. It is proved that the algorithm in this paper is more suitable for missing trajectory feature extraction.

This paper takes the optical fiber sensing technology as the research object. Firstly, according to the requirements of the sensing layer of the power Internet of things, starting from the multi risk monitoring of cable (overhead line), such as external breakage, fire, icing and galloping, an on-line monitoring system for transmission line environment is proposed, and the composition and deployment mode of the monitoring system are introduced. Secondly, through the research on the distribution of optical fiber temperature field and stress field, combined with big data and artificial intelligence, the monitoring of important environmental parameters of the line is realized. Finally, taking different monitoring scenarios as examples, this paper expounds the different monitoring functions of the transmission line Internet of things environmental monitoring system, which provides a useful reference for the transmission line environmental monitoring.

To further promote the improvement of pedestrian re-identification performance, this paper studies the reid framework based on "reid-strong-baseline", and uses different optimization schemes to improve the network performance. Firstly, the study tests three kinds of loss: Softmax, triplet hard, and Softmax + triplet hard, to verify the Rank-1 performance obtained and which can achieve the best performance. Secondly, based on the prototype network obtained by applying Softmax + triplet hard loss, we utilize several optimization methods including data enhancement, learning rate optimization, sampling method, and Label smoothing. Then we study the effectiveness of these optimizations on the performance of the Baseline model and the degree of improvement. Finally, this paper studies the efficiency of different Backbone and network depths on the performance of pedestrian re-identification.

Emotion recognition plays an important role in the field of human-computer interaction. It can help robots understand human needs more accurately. However, the impact of noise signal and model architecture on accuracy has not been fully explored. In addition, individual datasets are often used for algorithm testing, making it challenging to ensure algorithm generalization. To address these issues, we explore the impact of noise and algorithms on emotion recognition tasks based on two datasets. First, we use SE-ResNet as an emotion recognition network, which guarantees the effectiveness of the algorithm through the attention mechanism and residual structure. Experiments show that SE-ResNet performs better than other classical convolutional neural networks, and it validates the effectiveness of the attention mechanism. Second, we verify that noise can cause the algorithm to lose accuracy by setting up experiments with or without noise. Besides that, we analyze noise’s effect for each emotion class by the confusion matrix. The results show that noise has the most significant impact on the recognition accuracy of natural emotion.

In the water conservancy teaching, there are many resistance factors when students learn water conservancy experiments, so that students cannot learn water conservancy experiments better, and online and offline experiments alone cannot learn better. Therefore, in view of the above problems, another Combined with the current trend, this study intends to take the seepage experiment of earth-rock dam as an example and innovate the teaching method through a combination of online and offline methods, so as to better make up for the deficiencies in teaching and achieve a good teaching effect.

Aiming at the problem of low efficiency in obtaining flight boarding allowed nodes at large busy airports, the randomness and regularity of the flight from the actual time of arrival to the release of flight boarding allowed node is studied. After using Laplace feature map to reduce the data dimensionality, a prediction model of flight boarding allowed node is constructed based on support vector regression. The model analyzes and extracts the main factors that have an impact on the nodes that boarding allowed, and groups the daily data according to the airport's busyness to improve reliability. To improve the application effect, a historical database is established, and the purpose of dynamic prediction is achieved by matching historical data. The experimental results show that the accuracy of dynamic prediction is gradually improved. Within the error range of ±3min, the average maximum prediction accuracy can be up to 86.70%.

The performance of player object tracking in soccer match video is seriously affected by challenges like occlusion, out-of-view, similarity interference, low resolution, etc. To solve this problem, we propose a long-term tracking algorithm based on kernelized correlation filter. Firstly, the algorithm fuses shape, color and grayscale features to enhance the object representation ability and introduces scale filter to realize real-time scale estimation to improve the tracking accuracy and robustness. Secondly, the algorithm monitors tracking status in real time by the tracking result evaluation function. If the status judged good, the multi-peak re-detection of response map is used to review the tracking result. Otherwise, the object is re-detected through sliding windows to realize long-term tracking. The experimental results tested on Benchmark for Soccer Player Tracking (BSPT) demonstrate that the proposed tracker can achieve an accurate, real-time and long-term visual tracking for soccer player while runs at speed near 80 FPS.

In this paper, the identification system of sugarcane was studied based on the pre-seed cutting sugarcane planter developed in the laboratory to solve the problems of uneven cane discharging and high seed leakage rate. Firstly, simple structure analysis and system design of sugarcane planter are carried out. Secondly, histogram equalization algorithm is applied to enhance the image based on the real-time feedback of the camera. Thirdly, the template matching method is used to extract sugarcane images. Finally, the obtained sugarcane images were morphologic processed to obtain the surface texture information of sugarcane, and the information of sugarcane body and sowing situation were recorded through system recognition feedback.

In order to ensure the stable operation of the inter-satellite link network of navigation satellites, it is necessary to improve the monitoring and management capabilities of the navigation ground system for the inter-satellite link. In this paper, an inter-satellite link situational awareness system based on telemetry data is proposed and constructed. Aiming at the insufficient monitoring ability of the ground system for the inter-satellite link network status perception delay and the insufficient prediction ability of future network trends, the telemetry information extraction is carried out. output network status indicators. The grayscale prediction algorithm is used to predict the situation of various state indicators of the inter-satellite link network. The final result shows that the correct rate of telemetry count prediction difference within three in the state reaches 84.62%. It can effectively sense the status of future satellites and deal with the abnormal status of the inter-satellite link network in combination with other systems.

In this era of information explosion, we need to query through scholar website, talent website or major recruitment websites to find the required talent information. However, there are problems of easy matching failure, low correlation, high maintenance cost, complicated steps and lack of information. Considering with the current research direction and its short comings, this paper proposes a multi-feature and multi-relationship talent discovery algorithm based on knowledge graph (TDKG). Firstly, the talent graph is constructed based on talent dataset, then the needs of user are analyzed by natural language processing, and finally the multi-feature and multi-relationship search is realized by combining the talent graph. By crawling the real talent data on the post graduate enrollment information website, the talent graph and the talent discovery system is constructed for verification. The experiment shows that this algorithm can precisely identify the needs of users and return the talent information required by users. Compared with the existing talent search methods, it has more pertinence, richer and more perfect functions.

To address the fact that the traditional curve fitting method based on B-spline basis functions cannot preserve the sharp features in the original data well, a curve fitting method based on a class of orthogonal piecewise polynomial function Vsystem is proposed in this research. Firstly, different types of feature points from the original data are extracted by using the feature extraction algorithm; secondly, the feature points are reparametrized to the locations of different knots in the V-system; finally, the fitting curve is obtained by solving least-squares linear equations with constraints. Different features in the original data can be captured since the V-system contains basis functions with different smoothness. Numerical experimental results show that the proposed method in this research creates a fitted curve reflecting the global shape of the original data and can accurately represent sharp features.

Visual odometry is one of the key core technologies in the field of autonomous driving. However, images captured in lowlight or unevenly-illuminated scenes still cannot guarantee good performance due to low image contrast and lack of detail features. Therefore, we propose an end-to-end visual odometry method based on image fusion and FCNN-LSTM in the paper. The brightness image of the source image sequence is obtained by gray-scale transformation, and an image fusion algorithm based on spectral residual theory is designed to combine the image sequence and its brightness image to enhance the contrast of the image and provide more detailed information. In order to improve the accuracy of image feature extraction and reduce the error in the pose estimation process, we design a feature extraction algorithm based on skipfusion-FCNN. The traditional fully convolutional neural network (FCNN) is improved, a skip-fusion-FCNN network model is proposed, and three different paths are constructed for feature extraction. In each path, the prediction results of different depths are fused by downsampling to obtain a feature map. Merge three different feature maps to obtain feature fusion information, taking into account the structural information and detail information of the image. Experiments show that this algorithm is superior to the state-of-the-art algorithms.

Sarcasm detection aims to identify whether a text is sarcastic or not. In this paper, we propose a knowledge- and sentiment-enriched framework. Instead of modeling users' features or searching word pairs and snippets with sentiment conflicts in text, our framework integrates dialogue-related external knowledge and leverages inter-sentence sentiment to aid understanding sarcasm with the discussion context. Experiments on two discussion datasets show that our proposed framework yields better performance with enriched knowledge and sentiment information.

Ink and wash animation is a key component of animation art in China. China's ink and wash animation once fell into a stagnant stage due to the limitations of its production technology. However, with the rapid development of digital technology and the significant improvement of its manufacturing level, the development of ink and wash animation has ushered in new opportunities. The organic integration of ink and wash animation technology and art endows ink and wash animation with new appeal and vitality. More and more ink and wash animation creators began to use digital technology to create, such as using Adobe's Flash and After Effects joint drawing tools, Illustrator, and Photoshop to simulate various painting techniques of ink and wash painting, which improved the painting efficiency and quality. Based on this, this paper introduces the main expression ways of ink-wash animation and the bottleneck of traditional ink-wash animation development technology and analyzes the innovative significance and value of digital ink-wash animation technology. Taking the ink-wash animation of "Singing Goose" as an example, this paper discusses how to apply digital technology in ink-wash animation production, in order to provide reference and help for Chinese ink-wash animation producers.

Currently, 3D deep learning based on point cloud data has become a research hotspot in the field of computer vision. However, the high cost of acquiring point cloud data, the tedious process of processing and labeling, and the scarcity of high-quality and suitable datasets have been the prominent problems faced by researchers. In this paper, we propose a method to quickly produce point cloud dataset based on BIM 3D model and computer simulation technology, including the steps of classifying and labeling BIM models, converting 3D object data formats, extracting point clouds using Pytorch3d and Open3d libraries, and improving efficiency through Revit secondary development and Dos batch processing. Finally, we demonstrate the effectiveness of the method by performing semantic segmentation experiments using Pointnet++ network and analyzed the impact of point cloud sampling density, sampling method and 3D model accuracy on the performance of virtual point cloud. As a digital twin of the real world, BIM models are a natural database with rich scenes and all kinds of elements. It is hoped that the method studied in this paper can help researchers to produce datasets applicable to their own research and provide help for the application of 3D deep learning techniques in engineering and other fields.

With the development of The Times economy and the popularization of the Internet, digital technology and information technology have been rapid development, the improvement of people's living standards is no longer satisfied with the production of traditional clothing, but also can not meet the different needs of consumers. In recent years, as one of the most popular online shopping products, the two display platform can no longer meet people's requirements for clothing fabric, texture, fit performance, personalized customization and overall effect. According to the current situation of the virtual clothing design system. Through the comparative research of the existing garment virtual display platforms at home and abroad, the advantages and disadvantages of these platforms in the use process are summarized, and finally the development trend of digital and 3D virtual display technology in the garment industry is analyzed. The purpose is to clearly realize the current situation of the virtual design system, so as to put forward the corresponding solutions to the problems arising in the process, so as to accelerate the design process of the clothing virtual design system.

In network intrusion detection, using a machine learning method alone has blind spots and low detection accuracy. A stacked ensemble learning model using heterogeneous base-leaners for information security intrusion detection is proposed. Firstly, the convolution neural network is used to extract the deep information in the original data set, which is normalized as the input of the model. In constructing base classifiers, different heterogeneous model combinations are used to enhance the diversity of base classifiers. Experiments on NSL-KDD dataset show that the proposed model can comprehensively improve the detection accuracy, accuracy, recall and F1-score.

To address the scenario that there is the subjectivity of prior probability in the attack graph after the introduction of Bayesian network in the network attack model and the failure of attack nodes is not considered, an optimization scheme of the Bayesian attack graph and an intelligent construction method of attack path based on this scheme are proposed. The risk value of the target network is calculated to avoid the subjectivity of the prior probability and the devices are abstracted as attack graph nodes, and the atomic attacks are used as causal inference relations to reconstruct the attack graph. The analysis results show that the method has a significant improvement in the speed of attack graph and attack path generation and attack success rate, and it can perform the intelligent construction of attack path when the attack nodes fail.

Federated learning enables participants to construct a better model without sharing their private local data with each other. In the context of the continuous introduction of laws and regulations aimed at protecting data and privacy security, such as the "Data Security Law", federated learning has been more valued and more widely used. However, federated learning is vulnerable to attacks, one is backdoor attack. Here, we propose a backdoor attack method based on feature, used the CIFAR-10 data set and the ResNet18 model to research in the two different scenarios which one used the data that malicious participant participate in normal training as a backdoor and another used the data that implanting during the training as a backdoor. Especially, when we used the data that malicious participant participate in normal training as a backdoor, the attack success rate is about 50% while the attack does not affect the training process.

This paper analyzes the quantitative, qualitative and comprehensive risk assessment methods, introduces several reasonable ways to assess the target object, and establishes a hierarchical analysis model and an evaluation model based on the indicator system by setting reasonable risk factor indicators, and after setting the corresponding weights for the level of risk factors, the target network is evaluated and analyzed, and the security level of the network is finally determined through statistics and calculations.

In the era of big data, it is of great practical significance to improve the service quality of primary and secondary school libraries through digitization. On the basis of sufficient survey, an analytic hierarchy process (AHP) model is proposed to evaluate the service quality of primary and secondary school libraries. According to the proposed model, a library service evaluation index system for primary and secondary school libraries is constructed, including 6 first-level indicators, 16 second-level indicators and 28 third-level indicators. The proposed index system can comprehensively and objectively evaluate the service and management of primary and secondary school libraries, and has good application value.

This paper uses ARIMA-GARCH model to predict the prices of gold and bitcoin from 9/10/2016 to 9/11/2021, and fully analyzes the transaction date and constructs a price prediction model based on ARIMA-GARCH model, which can accurately predict the short-term price fluctuations in the future. When the transaction commission increases, the investor’s return decreases, but the gap between the investor’s return and the basic investment return gradually shortens. According to the official data in this thesis, the simulated transaction finally changed 1000 dollars into more than 9700 dollars, which performs well in the actual market.

Aiming at the threat assessment of unmanned aerial vehicle (UAV) cluster in formation flying stage, a threat assessment method of UAV cluster based on fuzzy analytic hierarchy process (FAHP) is proposed. Firstly, according to the characteristics of UAV cluster in formation flight stage, four suitable evaluation indexes are proposed; then, a threat assessment model of UAV cluster based on fuzzy analytic hierarchy process is established. Finally, an application example is given to verify the effectiveness of the proposed method.

In order to solve the automatic detection and alarm of indoor smokers, this paper designs and implements an indoor smoker detection and alarm system based on YOLOv5 model. This paper uses YOLOv5 as the target recognition model to complete the recognition of cigarettes and human objects, uses Deepsort target tracking algorithm to achieve the determination of smoking action, and combines Facenet face recognition algorithm and sklearn framework and other technologies to achieve face recognition and alarm function. The test experiment proves that the accuracy rate of the smoking person detection and alarm system based on the method of this paper is high and has certain practical and promotion prospects.

Traditional recommendation methods model users as vectors in a way that focuses only on single-sided user preferences. In order to compensate for the limitations of this modelling approach, a tensor modelling method is proposed that models the user as a rectangle. Firstly, a recommendation model based on a fusion of collaborative filtering and sequential recommender algorithms is constructed, which integrates the Transformer model and Pooling layer to model the user tensor; secondly, the similarity between the user tensor and the target item is calculated by combining the distance between the user tensor and the target item and the bias. The model is experimentally validated on the MovieLens datasets, and the results show that the model is able to focus on multiple user preferences and outperforms the baseline method in terms of accuracy of recommendation results.

Forecasting network traffic is critical to operators' daily network operating and maintenance decisions. However, nonlinearity, burst, and periodicity of network traffic, as well as the considerable geographical correlation among network nodes, pose significant hurdles to reliable network traffic forecast. Most existing traffic prediction methods use pre-defined graph or node embedding to extract spatial correlation between network nodes. However, neither of these methods may be able to extract this spatial correlation completely. Furthermore, while utilizing LSTM to extract temporal features, intermediate time step output is ignored, resulting in the loss of some temporal features. The network model in the article extracts the spatial correlation by the dual graphic attention module and extracts temporal features by the LSTM-Attention and TCN modules, which can extract spatial and temporal features from the network traffic data better. The network model is trained and predicted on the Abilene data set. The results demonstrate that the prediction performance is significantly improved.

Aiming at the problems of location and time limitations arising from the field visits in life, and then taking Changchun Institute of Technology as an example, a virtual campus roaming system based on Unity3D was developed to solve the problem of poor environmental simulation of traditional inspection methods. The system can improve the visual tension and expressiveness of users visiting the campus online, so that we can facilitate school publicity and promotion. First of all, the system uses 3DMax to accurately model the campus site, and then uses Unity to design and implement the functions, so that the system has the characteristics of high presence and strong interaction, thereby improving the user experience.

The PWARX model is a widely used modeling approach in hybrid systems. In this paper, a PWARX method based on LSTM is proposed. Firstly, the optimal control problem is clarified in the PWARX method. Secondly, the mechanism of PWARX based on the clustering method is described, and the LSTM algorithm is used to optimize the parameters and determine the region boundary conditions. Finally, the validity and reliability of the proposed method is verified by the engine universal characteristic curve, which solves the problem that the region boundary has gaps in the traditional method.

In order to understand the working state of on orbit satellites, it is necessary to analyze the telemetry data. The fast-changing telemetry data is an important data to express the navigation service status of navigation satellite. Its analysis and modeling are helpful to mine the deep information of navigation telemetry data. A modeling method of on orbit navigation satellite fast-changing telemetry data based on decision tree regression is proposed. The model is used to predict the power measurements at frequency points. The results show that R2 value is greater than 0.96, and the error of prediction value is small. A fast-changing telemetry data model with good effect is established, which provides a possible scheme for the application of artificial intelligence in the analysis of fast-changing telemetry data.

According to the national policy of rural revitalization, combined with the new technology of the current era, it has comprehensively opened a new pattern of rural construction, effectively combined with the strategic deployment of the Party Central Committee, and improved the deficiencies of some problems in Rural Revitalization. Through field investigation, we use virtual reality technology to highly simulate and restore the rural cultural heritage and its local characteristics. "VR add rural tourism" will be an important part of the Rural Revitalization Strategy. We should fully stimulate the field of rural culture, create tourism culture with Chinese characteristics, and make the countryside show its unique style and simple folk customs.

ZigBee network communication in the LAN is susceptible to interference from WiFi, Bluetooth, adjacent network obstacles, and other factors. ZigBee network itself cannot identify the type of interference, resulting in network reliability decline, and in severe cases, it will cause network paralysis. In response to this problem, this paper proposes a ZigBee network channel interference type identification method based on BP neural network, which accurately identifies the interference type by constructing neural networks in the ZigBee chip. If the decision output is WIFI or adjacent network interference on the same frequency, the ZigBee network communication channel is configured through the MAC layer of the ZigBee protocol stack to avoid interference. If the decision output is obstacle interference, the physical network can be reconstructed to avoid obstacles. Through simulation verification, the ZigBee network channel interference type identification method can significantly improve ZigBee network communication quality and anti-interference performance, which has application value.

Aiming at the complex recognition process of numbered musical notes in traditional methods and the low recognition accuracy, a method for recognition of musical notes based on convolutional neural network is proposed. Based on the structural characteristics of the notational notes, a single-input and three-output note recognition network model is established. The convolutional neural network is trained using sample images containing note information labels to obtain a note image recognition model. The training results show that the accuracy of the model and the change trend of the loss value perform well. In order to test the practicability of this method, the numbered musical notes of some songs were recognized and compared with other methods. The results show that this method has high recognition accuracy and fast recognition speed, which proves the validity and practicality of the recognition model.

Spectrum induced polarization method is mainly used for geological survey according to the difference of conductivity and polarization of medium. Combined with the needs of national strategic development, this paper studies the three-dimensional finite element numerical simulation method of complex resistivity based on generalized equivalent dielectric induced polarization (GEMTIP) model. This method has been widely used in resource exploration, engineering geology and other fields. First, the GEMTIP model and the complex resistivity variation characteristics of GEMTIP model under the influence of different parameters was introduced. Then, the variation equations were established for two-point sources of 3D modeling of complex resistivity method. The computing area was divided into hexahedral elements. The complex potential and the complex conductivity of rocks within each triangular lattice were described by a linear interpolation to create a linear equations system from the variation equation. The BICGSTAB (Bi-conjugate gradient stabilized method) algorithm with incomplete LU decomposition for preconditioning was used to solve the system linear equation to calculate the anomalous complex potential of all nodes and the apparent complex resistivity on the surface. Finally, this approach was verified through the calculations of a two layered model. Two typical geoelectric models were designed to test the correctness and efficiency of the algorithm. The results show that it provides new way to further study the induced polarization effect of rock and ore on the acroscopic scale.

With the promotion of smart city and other technologies, the application of embedded vision detection equipment is becoming more and more popular, among which gesture recognition is an important application of embedded vision detection equipment. At present, gesture recognition technology on embedded visual detection equipment is mostly implemented by calling API in domestic and foreign researches and products. But this method needs the support of stable communication network and has certain delay problem. To solve the above problems, this paper proposes a lightweight neural network model that can be deployed on embedded devices, which can realize local gesture recognition on embedded terminals without network remote transmission. The network builds a training framework on PyTorch and uses a homemade dataset for training, then lightens the network and finally deploys on raspberry PI for gesture recognition. Experimental results show that this network can run at a higher rate in raspberry PI 4B (4GB), and the model size is greatly reduced. The final recognition effect is good, and it has high practical value.

In order to generate an obstruction avoidance path for unmanned surface vehicle, the Dijkstra self-optimization model is proposed in this paper. First, a Maklink undirected graph is generated according to the starting point, ending point, and obstructions, the Dijkstra algorithm is used to generate an initial path to avoid obstructions. Second, a directed graph is generated based on the initial path, the Dijkstra algorithm is used to generate the optimal path with the shorter length. The experimental results show that the Dijkstra self-optimization model can generate the shortest path to avoid obstructions. Compared with the Maklink+ACO model, this model has smaller time complexity and is the optimal solution.

In order to effectively solve the problems of insufficient storage capacity, too late to shoot wonderful moments, and limited range of shooting space, a target video intelligent processing system based on Raspberry Pi is proposed. The system runs on the Raspberry Pi, and drives the camera to shoot a wider range of indoor environment video streams through the servo gimbal. For each frame of image in the video stream, image grayscale, filter denoising and histogram, equalization techniques are used for preprocessing. In the target detection and tracking stage, first use the OpenCV machine vision library to call the MobileNet lightweight convolutional neural network and SSD algorithm combined model (MobileNetSSD) for target detection, then it can alculate the relative position of the camera center focus and the target center, and finally drive the camera to track the target object. In terms of video processing, with the help of the results obtained during target detection, the automatic video editing process is completed by discarding the image frame without target object. Experiments show that the system can quickly and accurately track the target object to shoot, and effectively reduce the storage capacity of the video.

Traditional culture is the communication source and important communication channel of China's cultural construction output, which has a lot of value waiting to be excavated. With the rapid development of China's economy and technology, some excellent traditional culture has gradually disappeared in people's vision, many traditional folk crafts and technologies are on the verge of disappearing, their living environment is worrying [1]. The advent of the Internet era has brought new opportunities to the rise of traditional culture, which can spread its cultural deposits in different forms through the Internet platform. Under the background of current culture "going out", folk traditional culture has gradually attracted the public's attention. On the basis of respecting and restoring traditional culture, it is necessary to improve the effectiveness and interest of traditional culture information dissemination. This article mainly tells the story of folk traditional culture resources based on the cloud platform to optimize the design of the system, put forward his opinion in light of the present condition of the traditional culture, the purpose is to speed up the traditional culture and the integration of the Internet, thus the spread of the traditional culture and resources optimization design, as much as possible to retain the irreplaceability of traditional culture.

In view of the single control mode of traditional household appliances, which can't meet people's diversified needs of household appliances control, this paper designs a set of household appliances control system based on multiple modes. The system consists of three parts, including mobile app, host and slave. The master-slave control mode is designed with single-chip microcomputer as the control chip. The host receives the control command from the mobile app and sends it to the slave through the wireless module. After receiving the control command, the slave controls the household appliances through the relay module. Users can control household appliances in a variety of ways, such as button control, voice control and mobile phone control. After testing, the system is stable and reliable. It can meet people's demand for intelligent control of household appliances.

This paper takes the affricate of Tibetan Yushu dialect as the research object, and uses the acoustic experimental method to extract, analyze and summarize the acoustic parameters such as duration, center of gravity and dispersion, so as to describe the acoustic characteristics of the affricate of Yushu dialect. Firstly, the experiment uses the recording software Audition3.0 single channel adopts voice signal; Secondly, Matlab voice cutting program is used to generate a voice signal corresponding to a single name and voice; Then Praat is used to extract and analyze the parameters. It is concluded that the mean value of voice onset time can clearly distinguish the aspirated and unaspirated affricate of Yushu dialect; The more backward the pronunciation position is, the smaller the voice onset time duration is; The voice onset time value range of voiced affricate is larger than that of unvoiced affricate, and the voice onset time value range decreases with the back of the pronunciation part; The farther back the pronunciation part is, the longer the affricate time is; Compared with the unvoiced affricate, the center of gravity and dispersion of the voiced fricative spectrum are larger.

At present, taxi is one of the main transportation modes for passengers to arrive at their destination, and its convenience is increasingly becoming an important transportation mode in many cities. However, taxis queue up for passengers or passengers queue up to take the bus from time to time. Based on this problem, this paper establishes a corresponding mathematical model to analyze the taxi driver's selection strategy after sending passengers to the airport, so that the taxi driver can improve the efficiency of passengers as much as possible, so as to increase the available income. Firstly, in order to further put forward the selection strategy of taxi drivers, this paper reflects the driver's income according to the passenger mileage, no-load mileage, no-load rate and passenger carrying rate of taxis in different periods, and makes a selection strategy for taxi drivers from the aspect of maximum income. Finally, in order to consider the fairness of the airport taxi industry and make the income of these taxis as balanced as possible, we analyze and study from the two aspects of "short distance vehicles" and "long distance vehicles", and solve the problem of "boarding points" by setting up non passable lanes to improve the efficiency of the airport.

Gesture recognition, as an important means of human-computer interaction, can achieve more natural and flexible human-computer interaction, so it has been widely concerned by researchers in the field of computer vision. At present, most gesture recognition algorithms are based on monocular visual images and recognize the apparent features of hands. Most gesture image segmentation methods are carried out in color space according to skin color information. These methods are highly susceptible to interference from the external environment, such as lighting, background, etc. Convolutional neural network has the advantages of strong anti-interference and outstanding self-organization and self-learning ability. Therefore, based on the principle of convolutional neural network, a novel deep convolutional neural network dedicated to gesture recognition was designed in this paper. This network combines skin color information with finger position information for gesture recognition. Experimental results showed that the algorithm based on fingertip position information has better performance than the algorithm based solely on skin color information. Moreover, the network has simple structure and few parameters. Compared with VGG16 and other classical networks, the recognition accuracy is basically the same under the premise of fewer parameters and structural layers, and the recognition effect is better than other classical networks.

With the development of science and technology, virtual reality technology has gradually matured. The traditional method of promoting tourist attractions with the help of video cannot meet people's needs, so we build an ice and snow tourism system based on virtual reality technology. By software unity3D, build scenes, add characters, animals, snow scenes, etc. the virtual ice and snow world is present perfectly, provide people with a virtual space, strengthen people's perception of the ice and snow world by means of human-computer interaction, and let people experience ice and snow travel.

The development of digital intelligence technology drives the reconstruction of accounting functions and processes, which also puts forward new requirements and challenges for the training of accounting professionals. By analyzing the shortcomings of accounting majors in the context of the development of digital intelligence technology, and through the linkage of schools, industries and enterprises, this paper explores to build a new model of talent training reform driven by digital intelligence technologies such as Big data, artificial intelligence (AI) and blockchain, so as to promote the accounting education chain and talent chain to organically connect with the industrial chain and innovation chain, and meet the needs of the training and transformation of accounting professionals in the digital intelligence era.

Under the background of the epidemic in recent years, teachers and students are forced to carry out teaching online. Therefore, it is proposed to build an AVR virtual simulation network teaching platform. On this platform, Proteus Software is applied to simulate the hardware circuit controlled by AVR single chip microcomputer; Apply ICCAVR software to compile and download the program. With this method, teachers and students can complete the teaching and learning of AVR single chip microcomputer course online. Teachers can arrange personalized learning tasks on the platform level by level, and students can complete them one by one from easy to difficult. Teachers can help students solve problems in the simulation process through remote control. During the period of the epidemic, It turns out this new teaching mode has stimulated students' interest and improved their learning efficiency.

To address the error propagation problem of joint modeling of biomedical named entity recognition and normalization, joint label is designed to combine entity labels with concept labels to jointly label each term in the sentence, the joint learning task is transformed into a multiclass classification problem. A joint model of biomedical entity recognition and normalization labels based on self-attention is designed, the pre-training model BioBERT is used to encode the medical text. After extracting the joint label information using the self-attention mechanism, it is fused with the input sequence information. Finally, the final joint label representation is obtained by softmax. The experimental results show that the F1 values of the entity recognition and normalization tasks on the NCBI dataset reach 83.3% and 84.5%, and the F1 values on the BC5CDR dataset reach 84.2% and 86.6%, which are better compared with existing methods.

Under the high attention of the state, virtual reality technology has moved towards the public vision with its wide application ability. It has been widely concerned and praised by enterprises and users. Virtual reality technology has greatly reduced the tourism cost, solved some difficulties faced by traditional tourism, and brought new opportunities for tourism. This project establishes the virtual scene of Changbai Mountain in Jilin Province by virtual reality technology. Users can connect to the scene by virtual reality equipment and interact with the model in the scene to get a different tourism experience.

The development of digital intelligence technology has subverted the organizational management and business process of accounting, which makes the training mode of accounting talents in colleges and universities face challenges. Facing the needs of the digital intelligence era and guided by the outcome-based education concept, this paper has constructed a "One Body Four Wings" training mode for accounting professionals supported by the curriculum system, teaching methods, teaching staff and academic evaluation system, which has effectively solved the common problems in the training of accounting professionals in applied undergraduate colleges under the background of digital intelligence.

Driven by the dual carbon goals, China's carbon trading plays an increasingly significant role in carbon emission reduction. The price of carbon trading is the core issue of carbon trading, and predicting changes in the price of carbon trading is of great significance to the long-term low carbon development of enterprises. Based on an in-depth analysis of the properties of Prophet model and Long Short-Term Memory (LSTM) neural network, a combined Prophet-LSTM model for short-term prediction of carbon trading price is proposed in this paper. Prophet and LSTM prediction models are built separately, then a combination model is built by optimizing the weight coefficients for carbon trading price prediction. The data of Hubei carbon trading market from 2019 to 2022 were used as an arithmetic example for validation. The experimental results show that the combined Prophet-LSTM forecasting model has stronger stability and higher accuracy than the standard Prophet model forecasting method, LSTM model forecasting method, ARIMA model forecasting method and SARIMAX model forecasting method in carbon trading price time series analysis. The combined forecasting method proposed in this paper can effectively improve the prediction accuracy of carbon trading price, thus helping enterprises to reasonably assess their carbon assets and control the cost of carbon in production to achieve low carbon and high quality development.

The virtual reality technology is used to design the virtual reality system about ice and snow tourism to effectively present the viewing content in a sensory way, so that users can experience the snow atmosphere personally. It is the existence of virtual reality technology, multi - perception, interaction and other characteristics that make it popular with many people. In the future, virtual reality technology will be applied to all aspects of life and bring convenience to people's lives. This paper focuses on the research and construction at home and abroad, in order to understand the actual situation of VR, discusses the common problems in China, and proposes the strategy and application prospect of VR and its application in international development.

Buildings are commonly found as important facilities in today's society. How to detect building cracks safely and effectively is a necessary measure to ensure the safety of people's lives and properties. With the emergence of deep learning algorithms, various target detection methods based on convolutional neural network (CNN) models have gradually replaced conventional manual detection methods. In this paper, we design a crack recognition system based on convolutional neural network model for building images collected by UAVs. The experimental structure shows that the system has a good performance and can be further promoted to be applied in the field of safety assessment in the construction industry.

With the development of science and technology, the demand for automation and intelligence has nearly penetrated every corner of society. Single software and specific needs of information systems can no longer meet the growing needs of people. A complex information system composed of various systems, smart devices, and software emerged. The security of such complex information systems is becoming increasingly important. Attacks on complex information systems have become an important factor in harming national security, political stability, economic lifeline, and citizen security. Risk factors are weak links in the information system that may be threatened to cause damage, and the risk factors are transformed into damage to assets under certain conditions. Although the existing vulnerability management specification standards contain relevant content of risk assessment, the scope is not enough to support and cover the assessment of risk factors in information systems. In this paper, we comprehensively investigate and analyze the vulnerability standards of various vulnerability classification for information systems, and propose a classification standard for the analysis and grading of risk factors of complex information systems, which can provide a reference for the classification of information system risk factors in finance, public communications, and energy industries.

With the continuous development of mobile devices, the rapid increase in the number of Android malware poses a huge threat to malware detection systems. By classifying malware samples into families, the features shared by malware in the same family can be utilized in the malware detection method, to achieve the effect of improving the detection rate of malware. In this paper, a family classification method based on graph similarity is proposed, which constructs a family matrix and a weight matrix for malicious families and performs family classification by calculating the similarity between the software and each family. Experiments show that the classification accuracy rate of this method for the Kmin family, Inconosys family, Ginimi family, and DroidKungFu family in the Drebin dataset is over 90%.

As companies grow, the company's talent is growing, its projects are increasing, and its information assets are growing, it becomes essential to protect information security. This paper explains and analyses the security issues by using qualitative analysis approach and gives information security management solutions to protect information assets. By redesigning the company structure, assessing security risks, generating information system security management plan and establishing information security management system.

In this paper, a browser extension called “denglu1-extension” is presented, which helps users generate strong passwords and save them to the trusted user agent. The user agent can send the stored passwords to the browser extension to complete automatic login. Furthermore, in order to enhance the user experience, we also developed a captcha recognition module. According to experimental evaluation, denglu1-extension can efficiently avoid password vulnerabilities.

With the application of deep learning in object detection and recognition, remote sensing image recognition has also developed rapidly. However, the existing remote sensing image datasets are generally small, and remote sensing images are abundant and messy, so we use GAN to generate fake images that can be confused with real ones and build on this by using cGAN to expand the remote sensing datasets. At the same time, the existing research does not pay enough attention to low-resolution image recognition. We use the high-resolution model learning and the extended datasets to train classifiers, which can learn more general features conducive to low-resolution image recognition to improve the accuracy of the model used for remote sensing image recognition. In particular, we prune the model to reduce the weight redundancy of the network structure. Experiments demonstrate that on the SAT-4 dataset, our work both presses model size and maintains high model accuracy while effectively identifying remote sensing images.

System level software need read and write hardware memory, and some hardware system with limited resources such as 5G communication terminal in distributed grid application could not provide enough memory space for specified system application minimum demand. Some current method such as hardware expansion or software compression, to some extent, take effect, but could not solve the problem. Base on contrast between internal and external memory and single task independence, this article propose a realtime replacement mechanism to face on the situation and solve the problem met in high speeding data transfer distribute gird function as usual between running memory and limited resources. Experiment show that it could support limited resources system to run memory-needing task, and running efficiency is 90.2% about to hardware expansion.

The development of new generation information technology service education and teaching is an important hotspot of global development. The cultivation of new pharmaceutical professionals in higher vocational colleges must grasp the opportunity of the development of new generation information technology and realize the innovation and breakthrough of the cultivation of new pharmaceutical professionals. Taking the characteristic course of Internet of things technology as an example, studying the functional orientation and effect evaluation of the new generation of information technology in the training of new pharmaceutical talents can provide reference for higher vocational colleges to realize the integration of the new generation of information technology into professional talent training.

Neurons are highly morphologically complex, the whole brain image is huge, and strong noise, discontinuous signals, and mutual interference of signals often appear in neural images. The above problems have greatly increased the difficulty of neuron morphological calculation and analysis, so neuron morphology computation and analysis is widely regarded as one of the most challenging computational tasks in computational neuroscience. This paper introduces 3D-segmentation-net, an end-to-end learning method that can automatically segment 3D neuron images from sparse annotations. In automated segmentation validation experiments, we achieved an average IoU of 0.86. The network was trained from scratch and has not been optimized for this application. It is suitable for any mouse brain image segmentation task, and realizes automatic segmentation, tracking, fusion and real-time manual revision of a series of tracking schemes for massive neural images.

CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) is widely used in major mainstream websites as a security mechanism to classify human and computer. In recent years, the design and breaking technology of CAPTCHA have become an important research issue in order to verify security and reliability, which involves image processing, pattern recognition, artificial intelligence, computer vision and etc. This paper outlines some typical CAPTCHAs and presents an approach to recognize the text-based CAPTCHAs based on image processing and support vector machine algorithms. By testing the simple test-based CAPTCHAs, the experimental results show that the approach has an efficient recognition.

The time-domain wideband beamforming algorithm can output continuous waveforms in the time domain, which can improve the processing performance of array system. The time-domain wideband beamforming algorithm of the Frost structure needs to compensation the phase of the incident signal to the same angle through pre-delay. This paper proposes a robust Frost time-domain wideband beamforming method based on phase compensation, which integrates the pre-delay into the coefficient solution of the filter, suppresses the interference by adjusting the main lobe, and constraining the width of the main lobe to increases the robustness of the beamformer, its effectiveness is proved by simulation experiments.

Researchers have begun to pay greater attention to anthropometric measures as technology advances, and measurement technology has switched from contact to non-contact measurement, with non-contact measurement technology increasingly being used in the apparel industry. This paper analyzes, compares, and summarizes 2D non-contact measurement methods. The individual methods of image acquisition, contour recognition, feature point extraction, and dimension fitting for 2D non-contact measurement are introduced. The non-contact body dimension measurement based on computer vision is proposed and initially applied to the body dimension measurement of automotive seats.

Strain and temperature are two main parameters in fiber grating system, while there are cross-sensitive problems. In order to obtain this two-parameter at the same time, and to reduce its influence, a test system based on combined FBG is designed. Based on different effect of different core diameter strain, a kind of FBG probe is studied. On the basis of calculating the relationship between wavelength offset and the strain and temperature, the correlation function is deduced, and the coupling mode of the probe and the coupling efficiency are analyzed. In the experiment, strain and temperature test were completed by using light source, transmission fiber, combined FBG and so on. The results show that the wavelength offset is basically the same when the FBG changes with temperature. However, the slope change is quite different when stress changes. Finally, the strain and temperature were tested under different conditions, and the response coefficients of FBG temperature and strain were calibrated. It can be seen that the simultaneous analysis of temperature and strain can be achieved by detecting wavelength offset of combined FBG. The design is no need to weld, taper and other complex technologies. It has simple structure, strong capacity expansion capability and has high application value.

In order to realize the flexible construction and performance test of 4G communication system, in this paper, using the GPP (General Purpose Processor) and USRP as the hardware platform of Software Defined Radio (SDR), based on the open source srsRAN software platform, the 4G communication link is built which realizes UE accessing the Internet. Firstly, the peak rate of the system based on the RE algorithm is studied, and the influencing factors of the peak rate are analyzed. Secondly, the uplink and downlink peak rate of the 4G link is measured by the two transmission modes of antenna and coaxial cable respectively. After comparative analysis, it can be concluded that the transmission mode of coaxial cable is more stable and the SNR (Signal to Noise Ratio) is better. At the same time, it verifies that CPU performance affects uplink and downlink peak rates, and that adjusting the RF (Radio Frequency) gain can achieve the maximum throughput which is also consistent with the theoretical value of peak rates.

Music is a unique artistic form of expression. With the development of signal processing and data analyzing techniques, music visualization has been an important research issue in recent year. Visualization can enhance the music listening experience in a way such that an audience without any musical knowledge can correlate what is seen with what is heard and there by facilitate understanding the nature of music intuitively. This work aims at implementing a real time music visualization method based on Python which is one of the most popular programming languages. More specifically, we have implemented a system capable of visualizing music by analyzing the time and frequency information of the digital music and utilizing Python sound libraries. The results indicate that the system is generally successful and enriches listening experiences of the audience.

The evanescent wave that can only exist on the surface of the objects is the main reason for diffraction limit, which limits the subwavelength details of the objects to the near field. This paper designs a multi-layer phase gradient metasurface with 10°/mm increment and 69% transmission efficiency in X-band. Evanescent wave can be converted into propagation wave and transmitted to the far field with the help of this metasurface. Through Fourier transform and spectrum analysis, two targets with 10mm subwavelength distance are successfully resolved and imaged around 10GHz.

At present, navigation and positioning system based on location service plays an extremely important role in all aspects of people's production and life. GPS or Beidou navigation system performs well in outdoor positioning, however, satellite signals are difficult to meet the requirements of higher accuracy of indoor positioning. This paper designs a set of simple, low cost, high-positioning- accuracy indoor positioning system based on UWB, using DS-TWR ranging algorithm. Experimental results show that the system can achieve high precision positioning effect, the error is less than 10cm. The hardware design is completed, and the upper computer software is developed to draw real-time positioning map.

In the field of digital signal analysis and processing, discrete spectrum analysis is always an inevitable research topic. Signal period estimation plays an important role in improving the accuracy of spectrum analysis. When we perform truncation analysis on the signal, it is inevitable that the signal will be truncated by non-integral period, resulting in spectrum leakage. In this paper, we study the spectrum leakage law, obtain the leakage law of spectrum leakage attenuating by Gaussian function approximately along the frequency direction, and propose the minimum energy leakage period estimation method based on autocorrelation. The experimental results show that the algorithm is simple and can estimate the period of the signal when the signal-to-noise ratio is -10dB, which has strong anti-noise characteristics.

At present, people attach great importance to information security issues, and the implementation of different security level protection systems continues to deepen. However, some internal networks still need to face various forms of information security issues in the actual trial process. Based on the above content, this paper studies the design of the technical operation and maintenance safety management and control system of fort-and-fort machine, analyzes the main contents of the system design. At the same time, this paper also summarizes the relevant experience, hoping to provide a reasonable reference for workers in the same field.

In this paper, the job responsibility of railway locomotive and rolling stock supervision managements and operators is investigated, the current situation and problems of supervision management informatization are analyzed. Through the above research, the supervision business process model under the new supervision management mode is established, and railway locomotive and rolling stock supervision management information system is designed and implemented. Meanwhile, the actual application of the system is described at the end of this paper. Though the application of the system, it can assist supervisors to find and solve problems at the first time and realize the supervision management of the whole process of product procurement, design, manufacturing, testing, handover, etc, which strengthens the source quality control for the railway locomotive and rolling stock.

To better achieve image correction effect, an image correction method based on corner point detection is proposed. In image preprocessing, firstly, image equalization is achieved based on the contrast limited adaptive histgram equalization to avoid the problems caused by illumination and suppress noise while maintaining details, and then adaptive threshold segmentation is performed using the OTSU to obtain the binarized image. In the corner point detection stage, the contours of the binarized image are extracted firstly and the closed contours are filled to avoid the independent contours from affecting the accuracy of region growth, then the center point of the image is used as the seed pixel for region growth, and finally the four corner points are calculated based on the linear contours of the growth region and Hough theory, where the accurate region growth result can avoid the influence of the background on the corner point detection. In the correction stage, the perspective matrix is calculated by the four corner points, and the image is corrected by the perspective transformation. The experiments show that the proposed method can accurately find the corner points of document images and achieve efficient correction.

For the problem of poor accuracy and integrity of multi-view stereo (MVS) reconstruction, we propose an efficient multi-view stereo network, which mainly studies feature extraction module and depth map optimization module. At first, we introduce adaptive aggregation module to use context-aware convolution and multi-scale aggregation to adaptively extract image features, which effectively improves the extraction accuracy of weak texture surface. Then, we introduce dynamic self-learning optimization module (DSL) to solve the problem of depth map over-smoothing. Experiments show that, compared with MVSNet, our network greatly improves the integrity of reconstruction, and does not increase memory and time overhead.

For global attackers with stronger attack capabilities, this paper proposes a source location privacy protection (FS-SLP) mechanism based on pseudo-source nodes. The mechanism uses the global view information of the network and the location information of the current source node to dynamically generate multiple evenly distributed pseudo-source nodes in the network. The pseudo-source node transmits fake data to the sink node when the source node transmits real data packets to the sink node. packets, which greatly increases the difficulty of traffic analysis for attackers; to further enhance the security of the source node, this mechanism generates a dynamically adjustable filtering confusion ring near the sink node, and the nodes in the filtering confusion ring reduce the transmission path. At the same time, the transmission path of true and false data packets is confused, which reduces the energy consumption of nodes near the sink node and further increases the analysis difficulty of the attacker. Simulation experiments show that, compared with the existing scheme, this mechanism greatly prolongs the security period of the network on the premise of increasing a small amount of energy consumption.

Current tampering detection methods pay more attention to natural content images. The research on tampering algorithms for certificate document images is relatively limited, but certificate document images are the most commonly tampered with images, and they cause great harm to society. Our work presents a method for detecting certificate-like image manipulation using the ASGC-Net network. To achieve a network that can better localize text tampering cues. In addition, we propose a spatially constrained convolution that can effectively suppress image content and learn manipulation detection features by capturing different features between the neighborhood and the center of the convolution space. To increase the network's ability to capture tampering cues at multiple scales of images, we add multilayer cross-scale connections inspired by FPN networks. Experiments show that the algorithm is more accurate than general-purpose manipulation detection algorithms in locating tampered regions of certificate document images.

Deep learning technology has yielded good results in remote sensing image recognition of vehicles, but most existing recognition network models have poor interpretability, which limits its wide application. In order to achieve effective detection and recognition of vehicles in the complex environment, in this paper, the YOLOv4 is adopted to realize remote sensing images for vehicle target recognition. In addition, the optimized interpretation method with LIME is used to interpret the recognition results, improving the credibility of the recognition results.

In order to solve the problems in the programming of traditional marine CAD parametric drawing system, such as heavy workload of development and debugging, long period of modification and adjustment, and weak universality of the system, this paper adopts the development concept of configuration, uses the existing drawing files to quickly construct the command stream template file of parametric drawing, and automatically constructs the parameter input interface of the parametric drawing system based on the template file A more general CAD parametric drawing system is designed and implemented. The application of the method in the actual ship shafting design project shows that the method can quickly construct the CAD parametric drawing system of characteristic objects and improve the efficiency of related work.

In this paper, a scheme of laser-driven white light illumination based on diffuse reflection is proposed, which can effectively avoid the blue spot phenomenon. The white light with color temperature of 4717 K, lumen of 9.705 lm and color rendering index of 62 can be obtained by using the illumination module with 0.05 W blue laser excitation, which is consistent with the theoretical calculation results. We propose two methods to improve the white light uniformity, namely hook face and concentration gradient.

Large number of audio recordings are used in law enforcement and litigation procedures, and it also brings security issues such as the identification of the audio source. This paper mainly studies the problem of source identification (device detection). We proposed an audio source identification framework based on an improved residual network model that introduces a character category output, which will help to improve the identification accuracy for the special case of cross speaker. Experiments show that this audio source identification framework based on residual network has achieved good results under the condition of non-target recognition task, with the highest accuracy rate reaching above 98%, which outperforms the current audio source identification algorithm.

Cat and dog recognition is a classic problem in the field of image recognition. This paper proposes a fast detection algorithm FAST-CD-Classification-Net for this problem. The algorithm improves F1 from 0.7299 to 0.9875 with the help of residual structure; The product acceleration network reduces the running time from 0.025s to 0.008s without significantly reducing the accuracy. Experiments on the data set CD-KAGGLE show that the accuracy and robustness of the recognition algorithm designed in this paper are better than other algorithms.