This PDF file contains the front matter associated with SPIE Proceedings Volume 13078, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

A text detection algorithm based on attention-mechanism based feature fusion and enhancement for calligraphy fonts is proposed for the problem of diverse fonts, styles, forms, and complex structures in calligraphy works, and the network performance is improved by building an attention-mechanism based feature fusion module (AFFM) and an attention-mechanism based feature enhancement module (AFEM). First, according to the characteristics of calligraphic text, the AFFM module adds a feature pyramid FPN to combine global information with remote contextual information and improve the connection between each feature channel to improve the accuracy of text detection; second, the AFEM module improves the prediction accuracy of the model in predicting text pixels by enhancing the characterization ability of feature information after fusion of feature maps at different scales. The experimental results on the self-built text detection dataset of calligraphy works show that the precision, recall, and overall evaluation metrics F1 reach 94.29%, 94.07%, and 94.18%, respectively, proving the effectiveness of the method on the text detection task of calligraphy fonts. Meanwhile, on the public dataset ICDAR2015, its precision, recall, and overall evaluation metrics F1 reached 89.09%, 76.76%, and 82.47%, respectively, proving that the model has good generalization properties.

Low-voltage carrier reading system uses power line as transmission medium, and the power line has high loss and attenuation characteristics, so the signal is easily disturbed by noise during transmission, resulting in greater difficulty in transmission delay control. Therefore, a transmission delay control technology for low-voltage carrier aggregation system based on micro-power wireless communication is proposed. Support vector machine is used to extract micro power wireless communication features. Based on this, a delay predictor is designed to predict the delay of low-voltage carrier reading system by using the three-input, single-input and single-layer linear neuron structure of neural network. By using Hidden Markov model (HMMs), the current delay is estimated based on the transmission delay data of the previous time, and the transmission delay control of low-voltage carrier reading system is realized. The experimental results show that the throughput and fitting rate of the research method are high, and the delay can be significantly reduced under the application of the research method.

Various mainstream target tracking algorithms based on siamese networks are gradually becoming a trend in the field of deep learning tracking due to their concurrent advantages of accuracy and speed. Most siamese network-based trackers describe the tracking of a target object as a similar matching problem, and these trackers have achieved more advanced performance in several public tests. Most trackers often suffer from tracking drift or performance degradation owing to the non-updating of the template in the first frame and the target appearance encounters disturbing environments such as occlusion and drastic deformation. To address the current problem, this paper proposes an algorithm for fusion of multi-branch modules based on siamese network to achieve the fusion of target templates for updating, which improves the tracker's anti-jamming ability and the network structure is anchor-free. Meanwhile, this paper designs a new fusion module that fuses templates by inserting weight tensor method for multi-template fusion and optimises the results by complementary weight tensors. This method is practiced in SiamFC++ algorithm, where the target dataset is input and features are extracted, and then classification and regression operations are performed by fusion of multi-branch modules to get the predicted position of the target.

Simultaneous wireless information and power transfer (SWIPT) is an effective energy-saving technology, but its efficiency is hindered by environmental factors. The introduction of reconfigurable intelligent surfaces (RIS) has alleviated this issue, although it still faces significant constraints due to geographical limitations. This paper proposes a scheme that employs a simultaneously transmitting and reflecting (STAR)-RIS to assist SWIPT. It can overcome this limitation, achieve higher degrees of freedom (DoF), and provide better quality of service (QoS) for users on both sides of the ground. Meanwhile, we have considered a hybrid device based on a power splitting, which is capable of both energy harvesting and information decoding. We have proposed an efficient alternating optimization (AO) method to optimize the phase and amplitude vectors for reflection and transmission, beamforming and the optimal power splitting ratio, achieving an optimal balance between data rate and energy efficiency. Finally, simulation results demonstrate that the sum rate of the proposed model is superior to traditional RIS and other benchmark schemes.

Under the current climate change background, extreme weather events occur frequently, and meteorological factors will directly affect the quality of satellite communication links. In order to deal with the consequent risks, it is urgent to provide a satellite communication rainfall attenuation management system that can provide accurate monitoring and trend prediction. This will help to send a warning in advance, so that satellite stations can take timely response measures. Through the research on rainfall attenuation monitoring and analysis of satellite communication signals in Beijing West Satellite Station, a rainfall attenuation prediction model is established based on local historical and real-time rainfall data, the impact of rainfall attenuation on the channel in Ku band satellite communication is analyzed, and the solutions to effectively deal with the adverse impact of rainfall attenuation on Ku band satellite communication are proposed, Comprehensively improve the communication guarantee capability of satellite stations, and better provide stable and smooth satellite communication services for emergency communication.

Sample imbalance is an important problem in artificial intelligence. Datasets with too high imbalance ratio will cause the prediction performance degradation of the classification model. Many oversampling methods like SMOTE are proposed to solve the problem of sample imbalance. However due to the impact of noise samples and within imbalance distribution in datasets, performance of traditional oversampling methods were often limited. In this paper, we design a safety adaptive oversampling method combined with radial-based undersampling. Firstly, propose a region-partitioning criterion for the dataset, dividing it into safe region, boundary region, and noise region. Secondly, the undersampling strategy based on Gaussian radial basis function is combined with oversampling methods to remove noise samples and redundant majority samples that affect classification, and the number of redundant samples to be removed is determined based on the Pauta criterion. Afterwards, conventional synthetic samples are generated in the boundary region by a random linear interpolation strategy, while low variance synthetic samples are generated in the safe region by an affine shadowsampling strategy to reduce the offset of synthetic samples. Finally, in order to avoid the generation of noise samples, the deviation degree between each synthetic sample and their parent sample is discriminated by the concept of class potential, thereby ensuring a safer sample generation process. In experiments, 18 classic UCI imbalanced datasets are selected to compare proposed method with other oversampling methods. Experimental results show that our method has better performance than other listed oversampling methods.

Learning diverse detailed feature is crucial for fine-grained visual categorization (FGVC). However, most of existing methods for FGVC use the standard convolution for feature extraction which leads to the loss of many subtle but important features. Besides, in the existing attention models for FGVC, the features are aggregated by a simple global average pooling operation, which is unable to characterize complex feature information. To address these problems, we propose pyramid convolution and multi-frequency spatial attention (PCMFSA) for FGVC. To capture spatial context dependencies of different levels and avoid the loss of the subtle features, pyramid convolution, which uses convolution kernels with various sizes, is introduced in the backbone network. To reduce the influence of background noise and obtain diverse fine-grained feature information, we propose multi-frequency spatial attention (MFSA) module, in which two channel-wise grouping operations are designed and discrete cosine transform (DCT) in the frequency domain is adopted to enhance features of different spatial positions. Moreover, to mine more valuable regions, peak suppression is used to suppress high-response regions after MFSA. A large number of experimental results show that our method achieves state-of-the-art results on three popularly FGVC datasets.

Convolutional neural networks (CNN) are computationally intensive algorithms with rich application scenarios. In some applications, convolutional neural networks need to be deployed in embedded devices close to sensors. Field Programing Gate Arrays (FPGA) is highly favored in the research of convolutional neural network accelerators due to its design flexibility and low power consumption. Therefore, this article designs a convolutional neural network VGG16 based on FPGA. To reduce the area of hardware layout and design complexity, the design verification of VGG16 is completed on the Xilinx Zynq XC7020 development board and solved the problem of insufficient internal resources.

The Data Packet Marking method is a classic and effective approach for Internet Protocol (IP) traceback, and it has two traditional implementations - Deterministic Packet Marking (DPM) and Probabilistic Packet Marking (PPM). However, these two methods have certain shortcomings: DPM appends routing information for each hop to the packet header to achieve single-packet traceback, which results in oversized packet headers and fragmentation issues. PPM, on the other hand, includes only a single router's information in each marked packet, requiring a large number of marked packets for IP traceback. The IGPPM approach proposed in this article overcomes these limitations by marking multiple router interface numbers within a single packet. This allows recording multiple routing information within a single packet, effectively reducing the number of packets and the time required for traceback. Furthermore, it doesn't increase packet header size, avoiding fragmentation problems, and enhancing the efficiency of traceback.

This paper tackles the problem of network structure redundancy in image super-resolution (SR) algorithms and aims to develop an efficient and low-parameter SR method. It proposes a lightweight SR network based on vast-receptive-field information distillation (VIDN), which enhances the model performance and reduces the parameter size by introducing large receptive field convolution and attention modules, optimizing convolution operations, and using simple gate functions instead of RELU activation functions. The network leverages large kernel convolution and multi-scale attention mechanisms to better capture global information and fuse local and global features of the image, respectively. The ESA and CCA modules complement each other to improve image clarity and realism, forming a powerful multi-scale attention mechanism. The experimental results on three scales and four benchmark datasets demonstrate that the proposed algorithm achieves an average PSNR improvement of 0.03db, an average parameter reduction of 53%, and clearer and more natural visual effects on images.

Accurate seismic event classifications offer crucial information to earthquake alarms, and precise classifications in time can effectively reduce casualties and losses in seismic events. This paper proposes a classification model using a Vision Transformer network trained and tested using 8667 time-series records intercepted from seismic observatories. Different forms of input, including Gramian angular field and short-time Fourier transform, are used as the image input, and the regular time-series input is also tested. The precisions of classifications of explosions, collapses, and natural earthquakes are 93.12%, 92.49%, and 92.63%, respectively. The overall accuracy of classification is 92.74%, with short-time frequency maps of multiple channels of seismic signals as input, showing the ability of the proposed network model to extract and identify features of different types of seismic events.

Natural language processing technology refers to the use of mathematical methods and computer technology to model the form, pronunciation, semantics and other information of natural language, so that natural language can be formalized into one or more mathematical expressions. In the study of natural language processing, text summary generation technology belongs to an indispensable branch. This paper combines the TF-IDF algorithm, Textrank algorithm and LDA algorithm with keyword technology on the basis of the existing UNILM model to realize more efficient text summary generation.

This paper presents a BLE beacon-based Hospital Positioning System designed to improve the efficiency and accuracy of patient tracking and positioning in hospital environments. The system uses trilateration to provide real-time, precise positioning information by having each patient wear a BLE bracelet that transmits signals received by the BLE beacons. The signal information is collected and aggregated by the BLE gateway and processed on a cloud platform to determine the patient's specific location. The system also includes optimization algorithms, such as a greedy algorithm to determine the placement and number of BLE beacons, a shifting algorithm to smooth the RSSI values, and a trilateration algorithm to accurately calculate the patient's position. The proposed system has several advantages over traditional manual tracking methods, including improved accuracy, real-time tracking, and reduced workload for hospital staff. The system is scalable, flexible, and easy to deploy, and it can be customized to meet the specific needs of different hospitals. The system has been tested in a real hospital environment, and the results show that it is effective and reliable. This paper provides a new perspective and solution to improve healthcare services, especially patient tracking and positioning in hospital environments.

In order to deal with the problem of model gradients, neural network weights and other parameters exposing user privacy information and user privacy personalization under federated learning, therefore, this article proposes a personalized differential privacy method in a federated learning scenario. Before training, users upload the privacy protection level to the central server. The central server defines the amount of noise based on the total privacy budget and privacy protection level, and aggregates the training parameters using an aggregation algorithm weighted by the privacy protection level; the algorithm is verified to be effective in MNIST and CIFAR-10 data set and compared with existing localized differential privacy algorithms and centralized differential privacy algorithms, The classification accuracy is 94.12% and 45.20% respectively. Proved the usability and rationality of the algorithm in this scenario.

In view of the high demand for power communication network in the new power system services, this paper designs a power communication scheme based on the fusion of 5G network and SD-WAN. Firstly, the advantages of 5G network architecture and SD-WAN network architecture are analyzed respectively. Then, combined with the adaptability of new services, the above two technologies are deeply integrated to achieve point-to-point communication of power business terminals. At the same time, in order to ensure the power data transmission security, the network channel adopts the optimized SM4 algorithm, which meets national security requirements. Finally, the feasibility of the proposed scheme is verified by setting up a real network environment. The results show that the scheme proposed in this paper provides a strong support for the new power system communication network.

Visual simultaneous localisation and mapping is a fundamental technology in autonomous mobile robotic systems. The presence of dynamic objects in the environment can lead to incorrect feature matching, and factors such as fluctuating external lighting conditions can introduce instability to the system, thus limiting the practical application of VSLAM. In this paper, a robust VSLAM system for dynamic environments is proposed. Based on ORB-SLAM2, we use YOLO5 to enhance the consistency of the front-end combined with optical flow motion detection to detect dynamic targets in the environment and reject their feature points. Superpoint replaces ORB in the feature extraction process, which further enhances the adaptability of the system to the instability of the external environment. From the experimental results, it can be seen that the improved VSLAM, the real motion trajectory is extremely close to the estimated trajectory, and the error is greatly reduced.

In the field of mobile robot control, the utilization of reinforcement learning methods often faces the challenge of sparse rewards, resulting in suboptimal control performance. This paper proposes an approach that leverages the Diversity is All You Need (DIAYN) framework to dynamically generate reward functions and enhance the policy network weights of reinforcement learning algorithms. A comparative analysis is conducted with traditional reinforcement learning algorithms, namely Deep Deterministic Policy Gradient (DDPG), Deep Q-Network (DQN), and behavior-based robotic and proportional-integral-derivative (BBR PID) planning control algorithm. The results obtained from the CoppeliaSim simulation experiment demonstrate that, under identical training conditions, the DIAYN-based DDPG algorithm exhibits superior learning capability and achieves faster convergence to optimal actions. Consequently, the mobile robot can reach the target point more efficiently and with greater stability.

Superheat is one of the key indicators for reflecting production efficiency in aluminum electrolysis industry. To address the problem of insufficient performance in superheat identification with single modal data, a multi-modal deep feature fusion (MMDFF) model is proposed for superheat identification. By considering the global and local features synchronously, a cross-modal information interaction block is developed to improve the accuracy of superheat identification, which enriches the information expression of each mode. Finally, the effectiveness of the proposed superheat identification is demonstrated by the comparison experiments conducted on multi-source datasets of aluminum electrolysis industrial process. Meanwhile, the ablation experiments are used to verify the superiority of the cross-modal information interaction block.

This paper devises a decision-making method to address the varying communication requirements across diverse scenarios. The proposed method can objectively measure the performance of communication, and the weights of various indicators can be customized to adapt to different communication scenarios. Furthermore, in order to enhance communication performance, this paper has investigated a reinforcement learning models suitable for waveform and power decision-making, and proposes an explored action deletion algorithm based on the Q-learning method. The simulation results demonstrate that the proposed method outperforms existing methods in terms of convergence speed and algorithm accuracy.

Communication system failures often occur and it is difficult to find a certain pattern, this paper addresses the difficulty of fault prediction in electronic information systems represented by communication systems. Considering the characteristics of communication systems, such as complex structure, high integration, working principle and fault mechanism, this paper proposes to use the Multivariate Growth Model (MGM) for fault prediction based on the data, utilizing the Grey System Theory. The proposed model covers multiple characteristic variables causing communication system faults, monitors and predicts the characteristic variables in the operation of the communication system based on time series, and takes the predicted values as the basis for fault prediction. A certain type of communication system RF module is taken as case study to analyze the performance of the proposed fault prediction model. Through the analysis of the failure mechanism and the actual failure cases of the RF module of the communication system, four characteristic parameters, frequency error, medium power, squelch sensitivity and maximum audio level, are selected The comparative analysis shows that the proposed MGM method has better prediction accuracy than the traditional mean GM (1, 1) model. Besides, the model can process and calculate data with the help of computer software, thus improving the application efficiency of multivariate MGM (1, m) prediction model. The method proposed in this paper provides a feasible solution for predicting the critical faults of complex communication systems.

Intelligent driving technology is an important direction of the current development of automobile technology. The realization of intelligent technology is based on the perception ability of car sensors such as cameras and lidar for the driving environment detection. In the actual driving, the surface pollution of sensors will reduce the safety of intelligent driving vehicles. Therefore, it is very important to realize the automatic recognition of ultra-near-field dirt on the optical surface of car sensors. At present, the research on camera sensor fouling recognition is very scarce, and the relevant public data set for recognition research is also lacking. This paper conducted an in-depth study on the contamination problem of the camera sensor of the autonomous vehicle, independently created a dataset of dirt on the camera sensor, and trained the dataset based on the YOLO-V5 model using this dataset. The results show that the dataset we constructed is quite reasonable, with obvious features and excellent training results, with mAP(0.5) reaching 0.987; The results of the detection experiment prove that the trained dirt recognition model has certain universality, and has relatively excellent detection effect on different datasets, and the accuracy rate and recall rate can reach more than 0.8. This work could provide a certain reference for the cleaning research of automatic driving camera sensor.

Unmanned Aerial Vehicle (UAV) ad-hoc networks, characterized by distributed and dynamic topologies, are susceptible to interference attacks. This study delves into strategies to enhance the reliability and performance of UAV networks, with a specific focus on anti-interference measures at the Medium Access Control (MAC) layer. Two primary interference channel selection methods, model-based and model-free, are discussed. Model-based methods rely on prior information to estimate system models and predict channel interference states. However, these methods necessitate an in-depth understanding of the interference environment, limiting adaptability. In contrast, model-free methods utilize reinforcement learning, interacting with the environment to acquire real-time channel information, offering superior adaptability. This paper introduces an intelligent anti-interference solution at the central node of UAV networks based on the Deep Q-Network (DQN) model. DQN, a reinforcement learning algorithm, is employed to learn optimal channel selection strategies, thereby enhancing overall anti-interference performance. Research findings indicate that the proposed intelligent anti-interference solution achieves a high level of accuracy in channel selection, demonstrating robust anti-interference performance.

With the increasing convenience of transportation in recent years, the daily passenger flow of the airport has increased significantly. This has put forward higher requirements for baggage management at airports. In order to ensure the airport manages checked baggage and helps passengers find their baggage easily and conveniently, this paper presents an airport baggage positioning system using a wireless communication system. First, signals of both UWB and Wi-Fi frequencies are transmitted through the transmitter on the luggage. Then, based on UL-TDOA technology, the specific location of the baggage is calculated by the time difference between the different base stations receiving the signal. The system designed in this paper has obtained good simulation accuracy through MATLAB simulation tests. Compared with other similar products or methods in the specific environment, by higher positioning accuracy and other advantages.

Convolutional neural networks require a large amount of computing resources and time to achieve progress and development in computer vision tasks. Wavelet transform can provide multi-resolution features of images. By using wavelet transform to preprocess the images in the training set, the main information of the images can be preserved. The processed images can then be used as input for the neural network, significantly reducing the training time. By comparing different wavelet bases and orders, it was found that Bior wavelets showed the best acceleration effect, and the training time was significantly reduced. If the complexity of the model is appropriately increased, the training accuracy can be improved while the training time is reduced by 44% compared with the original time.

A network structure based on a combination of attention mechanism and large convolution kernel is proposed to address the problem of low detection accuracy of traditional multi-task traffic scene detection algorithms, which is used to perform vehicle object detection, drivable area detection and lane line detection tasks. Firstly, this paper proposes a backbone network that fuses large kernel attention mechanism and ELAN structure to improve the overall recognition accuracy of the model. Secondly, by fusing large convolution kernel attention mechanism and multi-scale information interaction mechanism, this paper designs a segmentation enhancement module to improve the model perform lane line segmentation task. Finally, by combining the characteristics and advantages of Focal loss and Tversky loss, this paper combines them as the loss function of segmentation tasks to solve the problem of sample imbalance in segmentation. The experimental results show that on BDD100K dataset, compared with traditional convolutional neural network algorithms, the recall rate of object detection of the proposed network reaches 91.1%, the accuracy of lane line detection reaches 84.0%, and the mean intersection over union of drivable area detection reaches 92.7%, which are significantly improved. Meanwhile, the proposed algorithm achieves a detection speed of 154fps on RTX3090, meeting the real-time requirements.

The standard Slime Mould Algorithm has problems such as falling into local optimal traps, slow convergence speed and low precision. In order to improve the performance of the algorithm, a new Multi-strategy Fusion based Slime Mould Algorithm (MFSMA) was proposed. In MFSMA, slime mould population was initialized with singer chaotic mapping and evenly distributed in search space, the global search ability was improved by alternating between short distance search and occasionally longer distance walk with Levy-flight mechanism, and a nonlinear convergence factor was proposed to balance the exploration ability and development ability of the algorithm. The proposed algorithm was able to find more precise parameters in various optimization problems than the conventional one. During the test phase, comparison test was conducted on MFSMA and other three test functions. The results indicated that the MFSMA had better global search ability, faster convergence speed and higher solving accuracy.

The prediction of bearing temperature is of significant importance for optimizing the operation and ensuring the stability of hydroelectric units. Based on practical operational experience, we establish a correlated mapping of bearing temperature during the operation of hydroelectric units and the main factors influencing its variations. We introduce a Support Vector Regression (SVR) model and employ the Particle Swarm Optimization (PSO) algorithm to optimize the penalty coefficient and insensitive loss coefficient of the SVR model. This leads to the development of a PSO-SVR-based bearing temperature prediction model for hydroelectric units. We compare the prediction accuracy of this model with other models such as BP neural networks and SVR. The results indicate that the proposed method yields predictions that are closer to the actual values, effectively achieving intelligent prediction of bearing temperature for hydroelectric units.

The slow real-time and historical data processing speed and insufficient data storage capacity of power grid equipment operation have led to an increase in risk detection time and a decrease in accuracy., A risk detection method for power grid equipment operation based on deep q learning algorithm is proposed for this purpose. Utilize big data mining technology to deeply explore the potential correlation between operational risks and early warning of power grid equipment, and establish a strong correlation model for risk early warning data. Use deep Q-learning algorithm to automatically identify abnormal situations in the operation of power grid equipment and achieve risk detection of power grid equipment operation. The experimental results show that the proposed method can accurately identify and predict potential risks that may occur during the operation of power grid equipment, and greatly shorten the detection response time.

The conventional cleaning model of dual-carbon data fusion is more inclined to data cleaning, and there are mistakes in attribute judgment during data fusion, which leads to the problems of missing and duplication of the output cleaning data, which affects the application of dual-carbon data. Therefore, this paper studies the dual-carbon data fusion cleaning model based on Kafka cluster. Select the comprehensive attribute weights of the dual-carbon data fusion cleaning model, fuse the data with similar characteristics according to the characteristics of the data attributes, and find out the missing values in the process of repeated cleaning, so as to avoid the data fusion deviation. Based on Kafka cluster, the regularization term of dual-carbon data fusion cleaning target is constructed to reduce the dimension of dual-carbon data, and a regularization term is constructed on the basis of data fusion cleaning target function to make the fused dual-carbon data more meaningful. The comparison experiment shows that the cleaning performance of the model is better and it can be applied to real life.

Conventional real-time storage methods for mixed data of power grid multi-task scheduling mainly use ORACLE RAC cluster to convert the storage format of mixed data, which is easily affected by the high concurrent response of storage, resulting in poor storage hit effect. Therefore, it is necessary to explore new methods to support the design of real-time storage methods for mixed data of power grid multi-task scheduling. This study designed a real-time storage consensus mechanism. Based on this, the real-time data storage algorithm has been updated. The experimental results show that the designed real-time storage method of mixed data for power grid multi-task scheduling has good storage effect, reliability and certain application value, and has made certain contributions to improving the efficiency of power grid multi-task scheduling.

With the rapid development of the digital society, hardware security has become a pressing issue in information security. Although the Post-Quantum Cryptography (PQC) is known for its high security, it is still threatened by hardware Trojan attacks. In this paper, the Classic McEliece algorithm is applied to hardware implementation and a hardware Trojan detection method based on sensor chains for PQC Classic McEliece circuits is proposed. The deployment of the sensor chain consisting of 12 ring oscillators (ROs) on the PQC Classic McEliece circuits is precisely controlled by employing an incremental compilation technique. The validation experiments are performed on an Altera Cyclone-II EP2C35F672C8 FPGA development board, and the results show that the hardware Trojans at RO6 and RO12 are successfully detected using the normalized difference algorithm. The Support vector machine (SVM) algorithm achieves an accuracy of 72%, the logistic regression algorithm achieves an accuracy of 92%, and the K-means algorithm achieves an accuracy of 97%. These results strongly support the effectiveness and accuracy of the proposed hardware Trojan detection method based on sensor chains for the PQC Classic McEliece circuits, and also suggest that the method is applicable to the protection of other cryptographic circuits.

The research on network switching technology is of great significance for seamless roaming of various terminals in heterogeneous wireless networks (HWNs). A personalized vertical handover decision algorithm considering terminal preferences in high-density access environments is proposed to address the dynamic characteristics of the network and the diverse needs of users. Based on the characteristics of the network, the algorithm maximizes the advantages of each network during high-density access, connecting more terminals to available networks with large connections as much as possible. In addition, criteria such as energy consumption and cost are selected to ensure terminal preferences, and a multi criteria decision making (MCDM) problem is constructed. The Analytic Hierarchy Process (AHP) is used to establish a judgment matrix, Determine the weights of each judgment criterion and ultimately select the network through utility function decision-making. The test results show that compared with the AHP based vertical switching algorithm, this algorithm eliminates the burden of networks and comprehensively considers terminal usage preferences during the pre screening stage. It can more accurately, fairly, and stably select the target switching network, reduce blocking rates and switching times, and more effectively improve the quality of user experience.

As digital transformation continues to advance, the demand for real-time synchronization of data between different internal application systems within government or enterprises, as well as between government or enterprise systems and external information systems, is growing stronger. Due to the diverse database management software used by various information systems and variations in the network environments in which databases operate, it becomes challenging to flexibly achieve synchronization and sharing of data. Therefore, there is a need to design a method that supports elastic scaling and flexible cross-network data transmission for real-time synchronization of heterogeneous databases to address the issue of real-time data synchronization. This paper will elucidate a solution based on cloud-native architecture for achieving the synchronization of heterogeneous databases. It aims to enable flexible deployment and high availability of data synchronization, as well as the elastic scaling of data synchronization tasks, while addressing issues related to data transmission in complex network environments and the synchronization of heterogeneous databases. Initially, the background technologies upon which this solution is based will be introduced. Subsequently, the overall implementation approach and the implementation strategies for individual modules will be detailed. Finally, the functionality and system deployment methods implemented will be validated through experiments.

The Guangdong Provincial Archives stores a large volume of valuable historical original Customs archives which are enormous in number and volume, formal in format, various in types and genres, rich in content, providing precious value for such studies as those of Imperial Chinese and foreign exchanges. It has always been the focus of research in the academic world at home and abroad. This paper, characterized by the inputting and outputting relationship among the entities (proper names of people, places, titles, weights and measures) that are most closely related to the translation of Imperial Guangdong Maritime Customs archives, constructs the knowledge graph of Imperial Guangdong Maritime Customs archival translation. In this paper, representative subgraphs being calculated based on the given node sets, a novel archival translation knowledge graph method is proposed, and the spatial components are integrated into the reinforcement learning framework to help understand the knowledge graph of archival translation. Through establishing the Imperial Guangdong Maritime Customs archival translation theory model, the internal structure and external information of archives are analyzed to obtain a more comprehensive and holistic view of the archival translation tasks. At last, the model proposed in this paper is proved to be reasonable by evaluating the data sets of Imperial Guangdong Maritime Customs archives.

The smoothness of high-speed railway tracks is an important indicator for judging the quality of railway tracks. Accurately predicting the uneven trend of the track and dealing with it in advance is of great significance to the safe operation of high-speed railways. According to the data characteristics of specific road sections, this paper proposes a SSA-LSTM model based on dual-stage decomposition to predict track irregularity of high-speed railways. First, in order to eliminate the influence of noise in the original data, Singular Spectrum Analysis (SSA) is used to decompose, denoise, and reconstruct the original data. Secondly, in view of the existence of multiple modal mixing effects in the denoised data, the data is decomposed through the ensemble empirical mode decomposition (EEMD) method. Then, to solve the problem that the long short-term memory network (LSTM) hyper-parameters are difficult to effectively determine, the Sparrow Search Algorithm (SSA) is used to search and optimize. Finally, the optimized LSTM model is used to predict and reintegrate the decomposed multiple sequences. Experimental results show that the SSA-LSTM prediction model based on dual-stage decomposition proposed in this article can obtain more accurate prediction results.

In recent years, with the continuous development and maturity of surveillance video technology, the complexity of the network environment, video surveillance identity security issues have become more and more prominent, and more and more surveillance video identity security issues are mentioned. How to ensure that the front-end device identity security access to the video surveillance system has become an urgent problem. In this paper, according to the "GB35114 public security video surveillance networking information security technical requirements" in the A-level standard requirements, designed a based on embedded MCU chip communication technology, to ensure that the front-end equipment legitimate access to video surveillance system security feasible program. The program uses the MCU chip to do data signature processing, data scheduling through DMA, and the front-end device authentication information with the SIP authentication server with security features for authentication. The method is low in complexity and can adapt to a variety of scenarios for signature verification. It promotes the further development of video surveillance security technology and applications, improves the security of video content, and thus protects the rights and interests of users.

Accurately predicting the performance outcomes of electrolytic cells holds paramount significance for subsequent production and decision-making within the aluminum electrolysis industry. Presently, the level of automation in electrolytic cell performance prediction methods is limited, heavily relying on manual assessments, which fail to accurately assess the stability of cell conditions and other related issues. This paper leverages knowledge graphs and graph convolutional neural networks for predicting electrolytic cell performance. Firstly, an electrolytic cell parameter knowledge graph is constructed to facilitate the structured processing of production data. Secondly, the graph convolutional neural network model is improved to enhance the performance of electrolyzer performance prediction by aggregating neighbourhood information through four layers of attention. Finally, validation is conducted through smelting experiments. Experimental results demonstrate that the proposed prediction method exhibits a high degree of accuracy, offering valuable insights for practical applications.

Aiming at the problems of the high dimension of features, high complexity of feature processing, and low efficiency of model detection of traditional industrial control network traffic data in complex network environments, this study uses an abnormal network flow identification and detection method based on random forest (RF), multi-head attention (ATT) and long short-term memory (LSTM) network. Firstly, the random forest algorithm is used to calculate the importance score of flow characteristics, screen out important features, and eliminate redundant features. Then, LSTM is adopted to identify and detect abnormal flows. In order to evaluate the effectiveness and superiority of the model, the accuracy, precision, recall, and F1-score are used in this study to evaluate the model, and the model is compared with traditional machine learning methods including Naive Bayes, QDA, and KNN algorithms. The experimental results show that the overall accuracy of abnormal flow identification reaches 99% on the CIC-IDS-2017 public data set.

The study of classification algorithms based on support vector machines is a popular research field. This article improves the deep multiple kernel learning model based on support vector machines by introducing fuzzy membership mapping, and proposes a basic framework for deep fuzzy multiple kernel classifiers, and improves the traditional fuzzy membership function based on class mean distance, so as to get a model of deep fuzzy multiple kernel classifier based on improved class mean distance fuzzy membership function. Experimental results show that our proposed model can achieve better classification results or stability relative to the original function on multiple datasets.

With the increasing complexity of computer programs, the number of reports of security vulnerabilities has also skyrocketed, making vulnerability detection a hot topic in the field of software security. Traditional security vulnerability research is based on manual processes or predefined rules. This kind of big data is underutilized. To solve the problems in the design of the above network model, in this paper, the Bi-GRU model is combined with the Attention mechanism to build a Bi-GRU-attention classification model. Bidirectional GRU model It contains two directional loop layers, each containing a gated loop unit, which allows the model to model the information in the sequence from both forward and reverse directions to better capture the dependencies in the sequence. However, for some long series data, the model will face a lot of interference and redundant information, attention mechanism can help the model to pay special attention to important information about certain elements in the sequence, therefore, adding attention mechanism can help the Bi-GRU model to extract and process key information more efficiently. Finally, the full connection layer and Softmax function are used to convert the results of attention weighting sum into probability output to complete the detection.

Based on the WWDB provided by Mantis, this article focuses on creating a large-scale, efficient, scalable, and realistic flight visual simulation environment. The method of developing large-scale multi region visual simulation systems using software was emphasized such as Terra Vista, Creator, Celerity, ect; And a detailed exploration was conducted on the LOD setting, texture mapping, region partitioning, and file conversion techniques during the development process. The methods for constructing a designated airport visual database on the basis of WWDB were summarized, and implemented seamlessly embedding of new visual database into WWDB. Enabling WWDB to embed multiple precision built visual databases according to requirements, meeting the requirements for visual databases in large-scale transition flight training of flight simulators, achieving the expected results.

This article presents a case study of our process in Smart Education. This is a case of detecting driver smoking behavior, in which intelligent image processing technology was used. Let students know how to conduct academic research through case analysis. Train students how to do image preprocessing, how to do detection algorithm, how to deal with experimental results, how to do result analysis and so on. Drivers have a kinds of common abnormal behaviors, including smoking, making phone calls, eating, and leaving the steering wheel. Among them, smoking is closely related to the fatigue state of the driver; when drivers make other abnormal behaviors, their hands will leave the steering wheel, which is easy to cause the car to get out of control. Therefore, this paper mainly focuses on the abnormal driving behaviors of smoking and the hands leaving the steering wheel.

Timely detection of abnormal behavior in power monitoring systems is crucial for system stability. Traditional log anomaly analysis methods have limitations, especially in handling multi-source or multi-dimensional log data. This paper introduces a method named CLSTMlog, which combines Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks, making the most of CNN's ability to extract local features and LSTM's capability to handle temporal relationships. Experimental results demonstrate that CLSTMlog exhibits a significant improvement in performance on the HDFS and IoT-23 datasets, including higher accuracy and reliability. This method has the potential to enhance the efficiency of log anomaly detection in power monitoring systems and overall system performance.