This PDF file contains the front matter associated with SPIE Proceedings Volume 12288, including the Title Page, Copyright information, Table of Contents, and Conference Committee Page.

To solve the problem of low rendering speed by drawing meaningless vertices when loading oblique photography model, a loading method based on image complexity is proposed. This method improves the nodal evaluation system in the traditional quadtree LOD loading method, and uses the image complexity of model node as the reference factor of the nodal evaluation system. The results show that compared with the traditional loading method which depends on the distance between viewpoint and model node, this method can improve the rendering efficiency and optimize the virtual roaming experience while maintaining the fidelity of the oblique photography model.

Vehicle re-identification (Re-ID) methods with supervised learning achieve high accuracy, but rely heavily on effective supervised labels, so that it cannot extend them to the unsupervised domain. Since the vehicle presents great changes from different perspectives and the large distribution gap between different datasets. Therefore, how to design a vehicle Re-ID method on label-free datasets and show outstanding performance is a difficult problem. In this paper, we propose an unsupervised vehicle Re-ID framework based on synthetic data. Our proposed framework consists of three steps: (1)we use synthetic data to generate pseudo-target samples similar in style to the target domain and use them for model pre-train; (2)the pre-train model is fine-tuned by the source and target domain to improve the cross-domain generalization of the model; (3)the orientation and the camera similarity are calculated by the pre-train orientation and the camera model of the synthetic data, thus punishing the final similarity. Experiments show that the proposed method outperforms existing stateof-the-art methods on benchmark datasets.

The placement of virtual network functions (VNFs) has been widely studied. We motivate our research by the following two observations. First, in many situations, such as enterprise and edge networks, the hosting servers may not have been deployed for VNF placement. Second, some VNFs may only process/cover a partial set of flows (such as elephant flows) but still improve network performance. To take the above observations into consideration, we propose to incrementally deploy servers and place VNFs with explicit performance guarantee. We formally define the Incremental Server and Virtual Network Function Co-Placement (InSCOPE) problem. We prove the NP-hardness of the InSCOPE problem. The results of extensive simulation show high performance of our proposed algorithm. For example, our proposed algorithm can reduce the number of servers by about 50%.

Time series analysis has played an important role in human production practice and scientific research. And the ARIMA model is a model often used when researching and forecasting time series. In this paper, we construct multiple groups of ARIMA models in the experiment and conduct white noise test and parameter significance test. The results show that the ARIMA (1,1,1) model is relatively optimal. Then applied the ARIMA model to describe and predict the PMI index of China's manufacturing industry, revealing the underlying economic laws. Judging from the forecast results of the manufacturing PMI index in year 2021: The ARIMA model can be used for short-term forecasting of the PMI index, and the model has a high degree of fit. The forecast model has certain application value. This model can bring better prediction and get more accurate future estimation in the field of robotics, automation and intelligent control.

Intelligent resource planning is expected to be of significant help in promoting the efficiency in tax halls, while service volume forecasting is a basis for further research on intelligent resource allocation and optimization. This paper uses the historical data from a tax hall in one of the most developed regions in China as dataset and studies the tax service volume forecasting based on time series forecasting, including five different models based on traditional and deep learning methods: ARIMA, Prophet, DeepAR, Transformer and the newly-proposed Informer. Aiming at selecting the best model for forecasting, the performance of the models on different business scenarios is compared. The experiment results indicate that the Informer has shown good performance and required reduced complexity in univariate forecasting. The results are expected to provide reference and further promote the research on intelligent task scheduling and resource allocation to enhance the service quality in the tax hall.

Deep learning is widely used in our daily life to solve some complex and tedious problems. In practical applications, if it was exploited by attackers would affect the reliability and security of the deep learning model. This article mainly introduces some attacks methods of generating adversarial examples from Generative Adversarial Networks(GAN) in recent years and related algorithms that use adversarial training to improve the robustness of deep learning models. At the end of the article, drawing on the reviewed literature, we present a broader outlook of adversarial attacks research direction.

In recent years, computer vision is more and more widely used in the field of intelligent control, and super-resolution reconstruction technology solves the problem of making the image clear. In order to solve the problem of insufficient image feature extraction and low information dissemination efficiency in the FSRCNN algorithm, this paper puts forward a dual-branch and residual network for image super-resolution reconstruction. The model designs a dual-branch feature extraction channel, expands the feature extraction channels, and improves the high-frequency information extraction ability of the input image; adopts an improved residual block to reduce the loss of information transmission. As shown in the experimental results, the peak signal-to-noise ratio (PSNR) of the Set5 dataset is 0.14dB and 0.52dB higher than that of the FSRCNN algorithm under the 2 and 3 scale factors, and the Set14 dataset is 0.13dB and 0.41dB higher respectively.

In the interdisciplinary field of finance and computing, scholars have proposed a method to calculate investor sentiment and thus predict stock trends from investor comment section data, which has also been transferred to perform analysis of the emotion of the news texts that would affect the investors’ decisions on investment. After the great success of textual sentiment analysis, some scholars found that the pictures’ emotions in the news have similar influences on investor sentiment. However, the problem is that, unlike the texts which can reflect the editor’s attitude directly and effectively, since some editors usually do not deliberately pay attention to the emotions of news pictures, most of them are published as how they look when taken. Few editors intentionally change pictures’ emotions according to their attitudes in post-processing moments. Thus, to apply a more precise stock forecasting model to automatic buying and selling software, it seems significant to figure out whether the news pictures and the texts express similar attitudes and in what kinds of news pictures can accurately convey the attitude of the news editor. We collected articles about finance from four sources and conducted a correlation analysis between textual sentiments and pictures’ emotions to solve these two problems. A modified NLP model provides textual sentiments, and a two-category deep learning model gives pictures’ emotions. On the whole, this work not only helps improve the accuracy of the automatic buying and selling software but also makes automatic control more intelligent.

This paper proposes an improved crack detection network, which can provide convenience in pavement maintenance and reduce the burden of labor costs. The network architecture is built based on YOLOv5-l, which has a balance of accuracy and speed. We introduce deformable convolution into the designed network to adapt to the crack target, which can improve the effect of detection. During the test, we also use test time augmentation to further improve the detection performance. The results on the RDDC2020 dataset show that the proposed model has reached a high level without using data augmentation and ensemble learning.

With the rapid development of mobile Internet, mobile terminals based on Android system are more and more widely used in industrial production and mobile office. One of the important application scenarios is remote access to the enterprise intranet for remote operation. How to access the enterprise intranet safely is a very important problem. Under the increasingly severe situation of network environment security at home and abroad, the application of domestic cryptographic algorithm is also imminent. Some domestic key fields, such as the State Grid, put forward the special needs of localization for network information transmission security. Therefore, it is very necessary to study the GMSSL VPN terminal based on Android. Firstly, this paper briefly introduces and analyzes the SM algorithm and GMSSL protocol standard. On this basis,introduces the key technologies on Android. Finally, the SSL VPN terminal based on Android system is designed and implemented, which adopts the SM algorithm and the GMSSL protocol standard. Compared with the international standard SSL / TLS protocol, it has better security. SM2 algorithm is used for key agreement, digital signature and SM2 certificate for identity authentication; SM3 algorithm is used to calculate message digest; combined with a variety of hardware encryption devices, hardware level data encryption is provided to realize a variety of ways of identity authentication, including encryption TF card, Bluetooth key, secure SIM card, etc. In addition, the security of terminal APK is analyzed and protective measures are taken to avoid the risk that Android applications are easy to be reverse cracked to a certain extent. Finally, through the experimental verification, the system also effectively provides security for remote terminal access on the basis of considering the ease of use. The research and implementation of GMSSL VPN terminal based on Android is of great significance for the popularization of Android mobile terminal in industrial production and the promotion of guomi algorithm.

News recommendation is one of the important tasks in recommendation tasks. The existing recommendation process is essentially a system loop feedback process. The content fed back by the recommendation system will be greatly affected by the user's click behavior, while the news recommendation model usually learns from the user's click behavior, which will lead to the system recommending news that users have clicked/liked to watch with high probability and pushing other types of news with low probability, resulting in the problem of insufficient diversity. To effectively address this type of problem, we first analyzed the news data statistically and found the bias of "viewing frequency", and then proposed a multi-perceptual news recommendation method based on dynamic multi-convolution kernel CNN and attention mechanism. We use CNN with convolution kernels of different sizes to extract headline features in the news feature extractor. And then combine them in an attentive way. Form title features with more representations. The user feature extractor is divided into biased user feature extractor and unbiased user feature extractor. One is to learn biased user attributes and perceive user characteristics biased and the other is to learn unbiased user attributes and perceive user characteristics without prejudice. Then we use the cosine similarity loss, that is, to increase the difference between the two user characteristics through cosine similarity. The most important point is that ensure that the accuracy of the model is not reduced as much as possible while debiasing, we combine the two features in an attentive manner, and finally propose a ranking method to improve the diversity of recommendations. Extensive experiments show that our method can not only remove biases, but also achieve a high accuracy of the model.

Application of simulation had proven successful in many different areas. In the real world, most systems are complex and huge systems, which need more complete models. It makes modeling work much more complicated and difficult. This paper presents a component-based modeling method and the model data organization of a discrete event simulation system (DESS). Component-based modeling method deals with the complexity when modeling an entity. In detail, we decompose an entity model according to the process of simulation which guarantees the reusing of component models. Then, we introduce a model data organization that mapping to the structure of component-based modeling method. In this case, we can compose models in an efficient way. Format of files and interactions between models also essential parts of integrating an entity model in DESS. Finally, in a simulation case, the process and data file of component models are presented to demonstrate the application of this data organization.

Based on the current situation that the equipment in the current distribution network system in our country can generally support the IEC61850 protocol, this paper studies the process of data conversion from the IEC60870-5-104 protocol and the Modbus protocol to 61850, and then realizes it by using the MQTT protocol The data can be viewed on the mobile phone. The hardware selects the Mir development board of the embedded Linux system. By changing the priority of the conversion process, the CPU occupation rate is increased, and the data conversion is completed on this development board. Then, while ensuring the security of data transmission, configure the MQTT server on the cloud server, and transplant related applications on the development board to upload data to the server. The mobile application can view the data in real time through the subscription server.

Knowledge reasoning is a key technology for constructing knowledge graphs and realizing knowledge graph completion. A single category of reasoning has its own advantages and disadvantages, so hybrid reasoning, which integrates the advantages of multiple single reasoning methods, has become a hot research direction. This paper will briefly introduce the background and definition of hybrid reasoning, introduce the classification of knowledge reasoning oriented to knowledge graph in detail, and then focus on the research progress, future research directions and possible challenges of hybrid reasoning in recent years.

Density peaks clustering(DPC) has been applied in the field of image segmentation. However, DPC only considers the global spatial information, but not the local spatial information. When a point is incorrectly allocated, the rest points will have a chain reaction, resulting in being divided into the wrong cluster. Therefore, we propose an image segmentation method based on superpixel and improved DPC. This method can fully consider local spatial information and is more robust. Firstly, an improved DPC is proposed to improve the clustering accuracy, and a new image segmentation algorithm is proposed by combining the improved DPC with superpixel. We compare this algorithm with other three algorithms in BSDS500 dataset. Four indicators of these algorithms show that the algorithm is superior to the other three algorithms.

Since the birth of BERT, Bert class models have achieved good results in Chinese Spelling Correction (CSC) tasks. However, the existing CSC models do not thoroughly learn typos' phonetic and visual knowledge. We propose a pretraining method and an Error-Prone Sentence Generation (EPSC) algorithm for training to overcome the shortcomings of existing CSC models. Experiments show that our method can improve the performance of BERT and soft-masked BERT in CSC tasks.

Erasure Coding is a advanced redundancy technique to ensure data reliability of the distributed storage system(DSS). However, when data is first written into DSS, it may be hesitantly to choose the specific parameters of erasure coding to balance the repair cost and storage utilization. In this paper, we design and implement VDRR-EC, a new distributed storage framework which exploit deep q network (DQN) mind to automatically decide the parameters of erasure coding to optimize traditional distributed storage system performance. Experimental results in real cluster environment show that VDRR-EC reduces data repair frequency and storage cost by 10.7%~19.2% and approximately 3.5% respectively, the penalty is acceptable additional network transport overhead and computing overhead.

Globally, the malaria epidemic remains serious, which poses an enormous threat to human health. Although China has made remarkable progress in the prevention and control of malaria, it still faces the potential for imported risks of this infectious disease. In this study, the SVR, GM(1,1), LR, and Exponential Smoothing models were applied respectively to predict the incidence of Malaria in mainland China. The results showed that the SVR model was the optimum model, and its prediction performance has higher than that of GM(1,1), LR, and Exponential Smoothing models.

Yellow rust is widely regarded as one of the most significant hazards faced by wheat. Traditional prevention methods are mainly based on pesticide application, which has high prevention costs and is harmful to the environment. Early disease detection and management can significantly reduce the cost of disease prevention and the impact on the environment. Therefore, a stable and reliable yellow rust detection method has become a direction for researchers. Spectral technology and deep learning methods are being introduced by more and more people to detect wheat yellow rust. In the recently years, Convolutional Neural Network (CNN) has got specific achievements of remote sensing image (RSI) processing. However, although CNN has achieved excellent performance, due to the locality of the convolution operation, it cannot learn the global and long-range semantic information interaction well. In this paper, we propose a dual flow transformer network (DF-Transformer), a combination of transformer and convolution, for multispectral image segmentation of wheat yellow rust. Tokenized image patches are delivered to a transformer-based encoder-decoder architecture with Unet blocks for global and long-range semantic feature learning. In particular, it uses a shifted window as an encoder, a maximum pooling-based U-Net block, and a hierarchical Swin transformer that extracts context functions. Next, an upconvolution-based decoder is designed to recover the spatial resolution of the feature maps. Experiments on multispectral images segmentation of wheat yellow rust demonstrate that the dual flow transformer network outperforms those methods with full-convolution.

With the needs of computer network secure storage, cloud computing technology continues to optimize and design the security model, so as to improve the overall performance of computer network secure storage and the anti attack ability of network storage system. There are also many problems in the improvement of computer network secure storage. Cloud computing technology designs and restructures the key parts to further maintain the stability of computer network secure storage. Therefore, this paper makes a comprehensive analysis of the key applications of cloud computing technology in computer network secure storage, so as to provide a development direction for the future development of computer network secure storage.

The problem of poor segmentation of diabetic retinal vascular images has always existed. In this paper, we propose an improved U-Net architecture RAFF-UNet to address it. In the U-Net encoding module, we innovatively use RAFF attention module before down-sampling. Based on the retinal vascular segmentation dataset made by Baidu PaddlePaddle AI STUDIO, the experimental results show that compared with U-Net and Attention-UNet, the MIoU index of RAFFUNet is increased by 0.94% and 1.33% respectively, and Acc index is improved.

The research purpose of this paper is to establish the graphic design system of embedded Linux technology, to meet the various needs of users and improve the product experience based on human-computer interaction, interactive design and user experience. In order to achieve this purpose, it is first necessary to use the system architecture design method, to establish the Microwindow system architecture, and to design the Framebuffer equipment driver, of course, but also to consider the needs of the system hardware circuit and interface design, to adjust the system according to the application of the system, to improve the effectiveness of the system design. The result is that the user experience design is fully satisfied, so it can be concluded that the computer technology can play a good role in the graphic design system.

Nowadays, the rapid development of new energy vehicles has also driven the new energy second-hand car market. This study aims to solve the current problems of messy residual value evaluation of new energy second-hand cars, and better predict the residual value through the loss variables. It crawled the data from the second-hand car website through crawlers and found that new energy second-hand cars have trends, periodicity, and some additional effects. Therefore, this study adopted the Prophet Multivariate Time Series model for prediction whose input is the licensing time and kilometers, and output is the predicted price. Through trend comparison and accuracy evaluation, the model showed better prediction performance whose prediction accuracy of the residual value of new energy second-hand vehicles can reach 96%.

In cut tobacco production line, the loosening and conditioning process is one of the most significant links affecting tobacco leaves quality. In order to solve the modeling difficulties of tobacco loosening and conditioning system due to time delay, strong coupling, nonlinearity and missing parameters, a data-driven model based on Long-Short-Term Memory networks is designed. Using the strong time series information learning ability and nonlinear fitting ability of the LSTM networks, it is trained only with the historical time series data of the outlet moisture and temperature of the loosening and conditioning cylinder, and the system model that can accurately predict the outlet moisture and temperature in output tobacco is obtained. The model predicts the output moisture and temperature values at the next time by inputting 60 consecutive historical output values. It is verified that the model has excellent fitting effect on both training set and verification set.

Birds have excellent three-dimensional space movement ability. The research and design of small and lightweight bird flight movement behavior information collection equipment has important application value in the research of bird robots and other aspects. Based on STM32F103 series chips, this paper designs a low cost small lightweight bird flight movement information acquisition device, and uses FREERTOS multi-task real-time operating system to realize the acquisition and storage of movement behavior information on this device.

As deep neural network models become more and more complex and the advent of massive amounts of data, distributed computing is needed to support the training process of the network model. Blockchain is a new distributed infrastructure and computing paradigm. Training the deep neural network model by converging the computing power of each node through blockchain technology will make the underlying consensus algorithm of the blockchain more practical and at the same time solve the problem of insufficient deep learning computing power. It studies the combination of blockchain and training of deep neural network model. It designs the consensus mechanism of the blockchain, uses the distributed training process of the deep neural network model as the proof of work of the nodes, encourages each node to contribute computing power through the incentive mechanism, and records the important data generated during the computing process in the blockchain.

Conventional crop fertilization classification and fertilization decision-making methods are difficult to operate efficiently and accurately. In order to improve the efficiency of greenhouse tomato fertilization and fertilization decision-making, this study is based on more than 1,800 greenhouse tomato leaf pictures, using data enhancement, migration learning, gradient descent, the regularization method optimizes the Convolutional Neural Network (CNN), and builds a deep learning model for tomato leaf fertility classification. The full convolutional neural network (FCN) is optimized using data enhancement and hyperparameter optimization methods, and a deep learning model for tomato leaf region segmentation and region information extraction is constructed. In this paper, the leaf image information collected by the on-site camera is converted into improved color space information, combined with the tomato leaf nitrogen deficiency-color model, to realize the amount of decision of the tomato nitrogen fertilizer at the irrigation site. Research shows that the accuracy of the CNN leaf fertility recognition model training set is close to 100%, the accuracy of the validation set can reach 95%, and the accuracy of the test set is 91%; the average intersection ratio from the test set data of the FCN leaf segmentation model is 0.91 and the average pixel accuracy is 0.94.

In order to improve the accuracy of human action recognition, a Residual Generative Adversarial Network (Res-GAN) method is proposed by combining the characteristics of Generative Adversarial Networks and Residual Networks. The generator uses an improved residual block extraction feature to obtain deep texture information and enhance the regularization of the model; residual blocks construct the discriminator. On the one hand, the accuracy of image feature extraction is improved, and the convergence speed of the network is accelerated. On the other hand, the discriminator performs the classification pre-training of human behavior recognition while distinguishing true and false samples. In order to avoid the influence of residual generative adversarial network generator on the recognition accuracy due to the instability of the reconstructed samples, the trained residual generative adversarial network discriminator was constructed as a classifier for human behavior recognition. Experiments on the KTH dataset demonstrate the effectiveness of the proposed method.

A multi-level HRNet-based algorithm for height-climbing behavior detection is proposed to address the problem of low recognition accuracy of existing height-climbing behavior detection algorithms. Firstly, the multi-level HRNet network is used as the human pose estimation module to obtain the 2D joint point data of the human body. Then the LSTM-based classification network is constructed, and the initial 2D joint point information is fed into the classification network for height-climbing action recognition of human pose. Finally, the target keypoint similarity OKS is used to evaluate the accuracy of keypoint recognition. Experiments were conducted on the COCO2017 dataset, and the accuracy of the proposed multi-level HRNet network for pose estimation in this paper reached 87.6%. Experiments on the human pose climbing action on the own dataset show that the accuracy of human pose climbing action detection of the multi-level HRNet and LSMT network proposed in this paper reaches 96.4%.

Aiming at the safety and efficiency of evacuation in fire environment, a 3d dynamic evacuation path planing method based on BIM techmnology and improved ant colony algorithm was proposed. Using BIM technology to build a parameterized model, as well as using fire dynamics software to simulate the model of the fire scene, and re-planning the evacuation path To solve the problem of insuficent inial pheromone of ant colony agorithm, the heuristic function was improved and the end point expectation was introduced.The experimental results show that this method can effectively improve the safety, ameliorate the initial path search efficiency of ant colony algorithmn,optimize the path length and make the fire evacuation path more intuitve,realize the three-dimensional dynamic evacuation path guidance of buildings, and ensure the safety of personnel.

In the design of distributed MIMO-SAR systems, the optimization of the satellite spacing along track is an important prerequisite for achieving the task of high resolution and wide swath. From the perspective of system performance optimization, this paper takes the along-track distance between adjacent satellites and PRF is used as an optimization variable, while considering the change of the vertical track baseline, combining genetic algorithm and nonlinear programming to obtain a global optimal solution. A method of optimizing the satellite spacing along track based on improved genetic algorithm is proposed and verified by simulation the correctness of the optimization method. The indicators and methods given in this article are of great significance for guiding the structural design of formation satellite systems and improving the performance of high resolution and wide swath measurement of SAR along track.

Tibetan-Chinese bidirectional translation can promote the development of Tibetan-Chinese scientific and cultural exchanges as well as educational and cultural undertakings, but it has been facing the dilemma of low resource language parallel corpus scarcity. In this paper, we hope to provide a research idea for Tibetan-Chinese machine translation under low-resource conditions through the study of pre-trained language models, and to alleviate the problems existing in Tibetan-Chinese machine translation such as scarce corpus and unsatisfactory translation performance, so as to promote the development of Tibetan society. In this paper, by introducing pre-trained language models BERT and ALBERT into Tibetan-Chinese bidirectional translation, the translation effect is greatly improved by using large-scale monolingual data. The translations before and after use are also analyzed. Through the tests, the average improvement of the Chinese-Tibetan translation is 2.13 BLEU values, and the average improvement of the Tibetan-Chinese translation is 2.68 BLEU values.

Artificial intelligence-driven design is a key topic discussed in the field of current art and design. For example, intelligent computer mapping belongs to mathematical modeling, in which to build artificial intelligence concepts and establish a cognitive science system, we can design computing models and carry out mechanical mapping. Through the combination of artificial intelligence and art design, the efficiency of art design can be effectively improved. Moreover, the essence of artificial intelligence is to simulate the human thinking through the artificial neural network. The circuit perceptron can make the law of artistic design clearer, and the establishment of an efficient model structure is the key to the combination of artificial intelligence and artistic design.

Most of the existing path-planning algorithms do not consider lateral safe distance requirements in practical applications. Hence, in this study, a new path point selection algorithm is proposed for path planning. The algorithm first used the Harris and Line Segment Detector(LSD) algorithms to detect and obtain the corner and edge information of obstacles. A vertical line was provided to the edge of the surrounding obstacles along each corner successively. In this process, the narrow impassable area in the map was filtered and removed by setting a safety threshold, and the foot of the vertical coordinates were simultaneously obtained. The corresponding midpoint coordinates were solved by using the corner coordinates and the foot of the corresponding perpendicular coordinates. The midpoint coordinates were used as candidate points to generate path points. These candidate points are screened and relaxed using the Probabilistic Roadmaps(PRM) algorithm to obtain the series of path points required. Finally, the path was planned according to these path points and smoothed using the Quadratic polynomial interpolation method(QPMI). Through simulation experiments, the method proposed in this study can solve a unique path without randomness under given conditions, and the probability of a collision in practical applications was reduced.

In this study, wheat at the filling stage of Hongzehu farm in Sihong County was used as the research object, and the wheat was monitored for head blight through an UAV equipped with a multi-spectral camera. The correlation analysis between 17 commonly used spectral vegetation indexes and wheat head blight disease index (DI) was carried out, and the 3 vegetation indexes with the highest correlation were obtained: normalize difference vegetation index (NDVI), structure insensitive pigment index (SIPI), triangle vegetation index (TVI), the wheat head blight monitoring model based on vegetation index was constructed and model evaluation was carried out. The results showed that the correlation between NDVI and DI value was the highest, and the coefficient of determination (R2 ) reached 0.8516. The stepwise regression equation model constructed with the three most correlated indexes as variables performed best. The coefficient of determination of model testing was 0.8787, and the root mean squared error (RMSE) was 4.1%. The comprehensive analysis shows that UAV multi-spectral remote sensing combined with the measured value of canopy wheat scab DI can realize the real-time monitoring of wheat scab in the field.

Due to the COVID-19 pandemic, many exams, written tests and interviews are conducted online and remotely, which raises a series of questions such as how to prevent cheating. In this project, the methods commonly used in the existing cheating monitoring system are fully investigated and their shortcomings are improved one by one. Finally, a line of sight detection algorithm based on computer vision technology is designed, and a prototype of auxiliary cheating detection system that can get good results only with a small number of samples is developed.

Medical image analysis is an interdisciplinary field of comprehensive medical imaging and analyzing, whose goal is to recognize disease diagnosis and lesion area through the related computer vision technology. Benefiting from the continuous development of the convolutional neural networks, medical image analysis based on deep learning has become a research hot spot. In this paper, based on in-depth literature research of results and progress in recent years, we mainly analyze the domestic and foreign research status of Medical Imaging in various application fields such as detection, segmentation and registration. We further compare the performance of representative methods on common data sets, and summary the existing challenges in deep learning-based medical image analysis. Finally, we discuss the solutions to these problems and predict the future development of medical image analysis tasks.

In order to address the issue that (scale invariant feature transformation,SIFT) algorithm will lose the original image color information, resulting in poor matching ability after graying the color image, this study proposes a SIFT image matching algorithm that incorporates color information . To get the grayscale image to be matched, the algorithm employs a bimodal distribution function to minimize the grayscale difference between pixels and the color contrast between pixels; On this basis, the nearest neighbor distance matching method using Hellinger distance as similarity measure and vector space cosine similarity matching method are combined for feature matching, so as to effectively maintain the color contrast of the original image, improve the number of effective feature points of adjacent images, eliminate the wrong matching point pairs and reduce the wrong matching. The algorithm can effectively discriminate regions with similar gray levels but distinct hues, enhance the matching accuracy, and preserve real-time performance, according to experimental results.

Negative emotion recognition plays a vital role in the field of affective computing. Recent studies reveal fuzzy boundaries among negative emotions that affect emotion recognition performance. However, to the best of our knowledge, there is no formal research to explore the impact of increasing negative emotions on the classification performance of emotion recognition. In this study, a novel dataset was designed and established with three sessions, which respectively include happy, calm fear (session1), session1 + sad (session2), and session 2+angry (session3). All analysis was based on differential entropy (DE) of EEG and heart rate variability (HRV) of ECG. The results show that the similarity of features of different emotions has increased significantly with more and more negative emotions added in the experiment process. This means that the increasing negative emotions video clips in the experimental paradigm affect emotion recognition performance to a great extent. The balanced types of emotional stimuli should be considered in the subsequent design for negative emotion recognition to avoid the adverse impact of increasing negative emotions.

Artificial intelligence technology is now widely utilized in a wide range of industries in China, with the merging of manufacturing and AI technology in particular deepening. AI technology, being a skill-biased technical innovation, is almost certain to have a significant influence on employment in China's industrial economy. This study constructs and quantifies China's manufacturing intelligence index using provincial panel data from 2010 to 2019, evaluates the mechanism of the impact of AI technology on China's manufacturing employment, tested empirically the impact of AI technology on China's manufacturing employment, and integrates the findings to draw conclusions on how China can achieve manufacturing employment in the AI technology environment.

Deep Reinforcement Learning (DRL) is poised to revolutionize the field of AI and represents a step towards general intelligence. Currently, AlphaStar achieved the Grandmaster level in StarCraft gaming, which is a remarkable breakthrough in Real-Time Strategy (RTS) gaming. The paper discusses the general framework of RTS gaming agents, and presents the method innovation from baseline RL algorithms to the AlphaStar. We begin with the basic idea and taxonomy of DRL, then progress to the technical framework of AlphaStar from the perspective of state space design, action space design, policy and value net framework and multi-agent training methods, presenting key issue and feasible method in full-length scale of RTS gaming. Finally, we draw a brief conclusion.

Aiming at the problem that the face image recognition technology based on anonymization mechanism cannot provide quantifiable privacy guarantee and realize controllable privacy protection, a controllable face identity recognition method based on differential privacy mechanism and generative learning model was proposed. Based on the attentional mechanism and graph convolution network, this method designed a representation learning model to extract identity-related geometric and demographic information by exploring spatial and spatial semantic relationships between different facial regions. The differential privacy mechanism was used to protect the demographic attributes while preserving the geometric attributes of the face. Then generate adversarial network reconstruction image instead of original image release for computer vision analysis task. Experimental results show that the proposed method is superior to other models in terms of privacy and utility. It can effectively resist background knowledge attacks, reconstruction attacks, and combination attacks while realizing verifiable and controllable privacy protection of face identity and ensuring the optimal tradeoff between privacy and utility of published images.

With the development of commercial 5G, the demands for indoor positioning with 5G signals are also increasing gradually. In this paper, we explore the commercial 5G signals for indoor positioning. Based on 5G smartphones available in the shelf, an Android application is developed to collect 5G New Radio (NR) signals and the fingerprinting methods of K-nearest neighbor (KNN) and Convolution Neural Network (CNN) have been investigated. The localization performance of 5G NR signals is tested in an office building. In the static experiments, the average positioning accuracy achieved by these two methods are 2.39m and 3.22m, respectively, while in the dynamic indoor tests, the average positioning accuracy is 3.53m and 3.08m. In both experiments, CNN performs better than KNN. The implementation shows that indoor localization by commercial 5G signals using existing 5G smartphones can meet the requirements of indoor office scenarios.

The agent routing process considering weather conditions is a multi-objective optimization problem constrainedly, which various optimization criteria and constraints should be considered. The quality of the given routes depends not only on the quality of the data, but also on the modelling of the optimization problem and the algorithm chosen to solve it. Because of universality of mathematical methods and the resulting decision-making challenges, the aim of this paper is to comprehensively compare the existing most advanced methods through literature review and elaborate different modelling methods, optimization algorithms and their application in agent ship route planning.

In text classification tasks, using traditional char granularity or word granularity information to encode context information into the same parameter space cannot effectively extract enough text feature information, and the expressive ability of text semantic information is also relatively weak. In order to solve this problem, this paper adopts the embedding method of fusing graphemes and morphemes, and establishes a text classification model that combines a hybrid self-attention mechanism and RCNN. The model first divides the text into char sequences and word sequences and fuses them, and then uses the Transformer encoder to extract preliminary features. The self-attention mechanism in the Transformer can better focus on semantic and location information, and then uses RCNN to extract the deep features which fusion grapheme and morpheme information, the feature vector is calculated and output by the softmax function, and finally the classification of Chinese short texts is realized. The experimental results show that the method proposed in this paper has different degrees of improvement in the accuracy of text classification compared with a single embedding method and a single neural network.

At present, the Covid-19 epidemic is still spreading globally. Although the domestic epidemic has been well controlled, the prevention and control of the epidemic must not be taken lightly. Being able to count the number of people in public places in real time has played a vital role in the prevention and control of the epidemic. Deep learning networks usually cannot be directly deployed on embedded devices with low computing power due to the huge amount of parameters of convolutional neural networks. This article is based on the YOLOv5 object detection algorithm and Jetson Nano embedded platform with TensorRT and C++ accelerating, it can realize the function of counting the number of people in the classroom, on the elevator entrance, and other scenes.

In the multi-objective optimization problem, it is an arduous task to select sub-populations with good convergence and diversity under the environment selection of evolutionary algorithms. Especially as the number of objectives increases, the number of non-dominated solutions will also increase sharply, so it becomes more challenging to increase the selection pressure on Pareto. To solve these problems, this paper proposes a hyperplane area selection evolutionary algorithm (HASEA). The algorithm constructs hyperplane area by the boundary points of individuals in the population and then projects all individuals to the corresponding area along the direction of the hyperplane orthogonal vector, and finally, each area selects corresponding individuals to form sub-populations through some strategies. Experimental results show that the subpopulations selected by this strategy have both convergence and diversity.

Expectations by computer vision technology to change traditional teaching model has become the research focus of many scholars, but due to lack of specific training data, and the classroom students intensive distribution, variety, easy to keep out each other, the existing multiple object detection and object tracking technique is applied to the poor performance of the teaching scene. For the above problems, in the stage of multi object detection, considering that the head can not only represent the multi-pose student object, but also is not easy to be completely blocked, this paper established a small dataset of teaching scenes with the head as the detection object to train and improve the accuracy of the Faster R-CNN detector. In the multi object tracking stage, the Kalman filter and Hungarian algorithm are used for tracking, and then the unmatched trajectory is remapped to the depth feature map based on the strong feature extraction ability of Faster R-CNN backbone neural network. Then, we correct the unmatched trajectory by fusing the depth feature information and historical trajectory position information of the object, and calculate the trajectory similarity, which improves the problems of identity switching and trajectory interruption caused by the change of object attitude and occlusion. The research of this paper is used in real teaching scenes, which can assist teachers to track students' status and the formed student object trajectory sequence, and can provide data input for subsequent classroom behavior analysis.

Panorama technology can build real virtual reality technology for users on the computer based on panoramic images. Panoramic technology can bring users an all-round interactive view of the restored display of the real scene. Panoramic technology can help the red station to promote, let people know more about the red station, and improve the service quality of the red station. This article introduces the application of panoramic technology in the design of red station, and deeply analyzes the panoramic scene to introduce the way the system integrates panoramic technology, so that users can observe the red station through a 360-degree panoramic view through smart devices.

Radar cross section(RCS), extrapolated from small angle rotation in high frequency band by the traditional inverse synthetic aperture radar(ISAR) imaging, can not be presented full aspect measurement. In order to address this problem, we propose ultra-band large angle rotation method to derive RCS from ISAR imaging. This method analyze the distribution characteristic with small angle rotation between radial range and cross range as well as illuminate the point spread function(PSF) of target scattering distribution mechanism under measurement and imaging condition. Experiment show that this method not only can be strengthen the scattering position image of target, but also enhance the RCS resolution.

Aiming at the problems that the melody generated by the existing multi track music generation model is not pleasant and the harmony between tracks is not enough, a multi track music generation model with TransformerX-SpanBERT sequence conditions is proposed. The module is a Generative Adversarial Network structure. Since the generated music has multiple tracks, first, use TransformerXL as the generator and SpanBERT as the discriminator to generate piano track music, and the other tracks have one generator for each track. Therefore, there are multiple generators and one discriminator. First, piano track music is as input, let each track generator to generate their own music. Then, the common discriminator is used to judge the music quality generated by each track and the harmony between tracks. Finally, the global attention mechanism is used to pay attention to the music characteristics to ensure the overall effect of music. The objective evaluation index of music and the subjective score of listeners are used to verify the model and compare it withTransformer-GAN and MuseGAN Multi-track sequential symbol music, the results show that the music generated by the model in this paper has the advantages of more harmonious tracks and better melody, rhythm and coherence, therefore, it can be applied to practical music creation.

The integration of artificial intelligence technology and tax risk management can establish an intelligent tax system, and through Bagging and SVM computer algorithms. Both an intelligent tax system and a predictive tax evasion model are essentially a tax system based on computer technology. The use of forecast indicators, tax cases and tax data can improve the accuracy of tax risk prediction on the basis of reducing tax cost, the accuracy of tax risk and further ensure the accuracy of tax data. Through this model, the control level of tax risk management can be improved.

The network traffic prediction is the basic premise of tasks such as dynamic planning and congestion control. However, due to the burstiness, volatility, and non-stationarity of the short-term network traffic, the prediction model is required to possess high timeliness and accuracy so that it is very difficult to predict. In order to solve the above-mentioned problems, an Improved Temporal Convolutional Network (ITCN) algorithm is proposed. This algorithm introduces a historical influence coefficient to improve the TCN algorithm, which heightens the efficiency of the TCN in processing traffic time series and simultaneously enhances the ability of the TCN algorithm to capture temporal features. The simulation experiments indicate as follows: in contrast to the comparison algorithm, the accuracy of the ITCN algorithm is improved by 60.99%; the training speed is improved by 94.52%, and the algorithm prediction speed is improved by 72.76%; the ITCN algorithm can meet the requirements of the accuracy and the timeliness for the short-term traffic prediction tasks.

Aiming at the poor effect of multi focus image intelligent fusion, a multi focus image intelligent fusion method based on improved generative adversarial network algorithm is proposed. The multi focus image recognition is carried out by collecting the multi focus image features, the multi focus image denoising is realized by using the improved generative adversarial network algorithm, and the multi focus image intelligent fusion model is constructed. Finally, the experiments show that the multi focus image intelligent fusion method based on the improved generative adversarial network algorithm has significantly improved the definition of multi focus image fusion in the process of practical application, Fully meet the research requirements.

The network attacks targeted at the power system become much more complex and covert. However, traditional security protection methods cannot well detect such unknown attacks or multi-step attacks, which leads to a constant threat. How to use the massive log and warning data generated from various types of traditional security equipment to effectively detect and trace the advanced network threats becomes more and more important. "Graph" is a data structure that can better represent the power network system and thus enforcing correlation analysis on the network graph is an important detection method. This paper first reviews the graph construction methods and graph representation learning method. Furthermore Dynamic-Graph Neural Network (DGNN) based method to detect attack behavior is proposed. The experimental evaluation shows that DGNN based method can achieve better performance compared with both shallow embedding method and static-GNN method.

In traditional collaborative filtering recommendation system, drawbacks such as interest drift and data sparseness always exist, which has weakened performance of the system. In this paper, a novel approach from the perspective of emotion analysis and forgotten rules was proposed to offset the influence on recommendation accuracy caused by interest migration. Firstly, user online comment has been analyzed to extract emotional words and the rated object using optimized DP algorithm. Secondly, the emotional words about the rated object have been appropriate quantified so that we can take advantage of the user emotional tendency to generate interest model. Thirdly, the Ebbinghaus forgetting curve is also introduced into the recommendation process to overcome the problem that caused by the change of user interest, which can effectively increase the accuracy and offer user better service when interest move. The simulation experiment shows that the strategy could greatly reduce MAE and effectively improve coverage.

Computer network has been widely used in all walks of life, network security has also received unprecedented attention, network intrusion detection technology is one of the key technologies to maintain network security. Traditional intrusion detection methods based on rules have some disadvantages, such as dependence on manual intervention, difficulty in updating rules database, and difficulty in detecting unknown intrusion. Therefore, a lightweight network intrusion detection method is designed based on improved federated learning. Firstly a network intrusion detection model is constructed based on improved federated learning to realize lightweight network intrusion detection. The experimental results show that the designed network intrusion detection method has good detection effect and has certain application value, which can be used as a reference for subsequent network intrusion detection.

We witnessed a lot of examples of cases launched by Advanced Persistent Threat (APT) organization and the spectacular destructive power. The realistic data originate from occurred attack events almost is not available. Fortunately, there is some simulation dataset. Many previous works were chosen to evaluate their method on the Defense Advanced Research Projects Agency (DARPA) Transparent Computing Engagement 3 dataset that contains attack behaviors implemented by a highly skilled team. Meanwhile, with the rapid update development of computer technology and the competition between detection means and attacking methods is becoming increasingly fierce, this dataset is outmoded, it exists the enormous gap between current cyber attacks and the DARPA TC E3 dataset. In this work, we provide the statistical information of DARPA TC E5 and compare it with DARPA TC E3. We analyze the practicality of the more advanced DARPA Transparent Computing Engagement 5 dataset.

With the continuous development of information technology, industrial control systems have become an importantpart of the national information infrastructure, and the network security risks are further complicated. It’s imperative to use new technologies and methods to detect and defend against attacks.This paper constructs the LightBGM-SVM model, tests it combined with the experimental data of industrial control system, and compares the differences of three common intrusion detection models in detection rate, false alarm rate and response time.The verificationresults show that the LightBGMSVM model is significantly better than the other two models in the test set verification. The application of this model to the intrusion detection of industrial control system has better accuracy, and further defines the optimization direction of intrusion detection for more complex industrial control system.

Mining low-resource multilingual parallel sentences would be useful in natural language processing tasks. A large number of high-quality parallel sentences provide necessary data support for tasks such as building bilingual parallel corpora and cross-language information retrieval. In this paper, we present an approach to mine Thai-English parallel sentences in huge documents using cross-lingual sentence embedding. To evaluate the approach, we used two extensive bilingual corpora, which provide golden scores. On TED and Tanzil set, our approach improves nearly 0.73 points in AUC and reaches 96.5%. In the task of mining BUCC parallel corpus, our approach uses less time and space but gets an F1 score similar to the LaBSE model, which has reached state-of-the-art on BUCC. Our model not only solves the problem of sentence alignment with insufficient resources but also uses less time.

In order to reduce the probability of vehicle rear end collision accident in foggy weather, and in accordance with the background of accelerating the development of automobile safety industry with artificial intelligence technology, a vehicle recognition, detection and voice early warning based on improved yolov5 algorithm is proposed. Firstly, adaptive convolution is introduced into convolution neural network to enhance the feature extraction ability of target in the process of training network model, and then improve the expression ability of neural network for important feature channels, especially for small target recognition and detection; Then, considering the real-time nature of detection and early warning, the neck detection layer in yolov5s network is cut and optimized to improve the reasoning speed of the network model. Finally, through the comparative analysis of the evaluation indexes such as the accuracy, recall and map of the training model, and the test of the scene pictures of the actual driver's perspective in foggy weather, it is verified that the improved yolov5 algorithm can improve the detection ability and reasoning speed of small target vehicles. The experimental results show that the accuracy of the improved yolov5 algorithm is as high as 97% and the reasoning speed is improved by 1/3. The driver can be warned in real time to reduce the probability of rear end collision accidents.

With the continuous development of deep learning, anomaly detection technology in computer vision has maderemarkable progress. Video anomaly detection is essential to ensure public safety. We classify and summarize anomalydetection based on deep learning. First, the overall process of anomaly detection is presented. Then, based on the neural network training method, we discuss the development and application of deep learning in the field of anomaly detectionfrom four aspects: Multiple Instance Learning, Regression models, Clustering models, and Reconstruction models. Finally, we present commonly used datasets and performance evaluation criteria , analyze the performance of different algorithms and discuss the future directions.

Relation extraction is the core mission and a great significant part of natural language processing. This paper briefly expound the development of relation extraction and introduces the commonly used document-level relation extraction datasets and evaluation indexes of model effects. According to the different representation of entity by model, documentlevel relation extraction can be divided into sequential method and graph method. Besides, we conduct contrastive analysis concerning different relation extraction models, and compared the effects of various relation extraction models. Finally, we summarize the key research contents of document-level relation extraction in the future and forecast the development trend.

Under partial shadow conditions, in the use of PSO algorithm in MPPT, there are problems such as slowastrictionspeed, likely to access to local optimization and low tracking accuracy. This paper put forward an improved particle swarmoptimization algorithm apply to maximum power point tracking control method. This method introduces a randominertia weight factor and adds mutation disturbance to the update of the particle population position to enhancetheconvergence speed and search ability of particles, and enables the "premature" particles to quickly jump out of thelocal optimization, thereby enhancing the optimization of the PSO algorithm performance. After simulation analysis, theresults display that compared with the PSO algorithm, the method proposed in the article has a faster convergence speedand higher tracking accuracy in the process of system tracking, thereby enhance the output efficiency of the PVarray.

Aiming at the problem of large amount of calculation and difficulty in convergence of robot path planning in complex underwater 3D environments, the paper proposes a 3D path planning algorithm based on the sparrow search algorithm with potential field heuristic. First, establish an underwater 3D grid model, and use artificial potential field forces to generate an initial feasible path. Secondly, use the infinite folding chaotic map to expand the sparrow population and increase the diversity of the population.Then use the dynamic adjustment strategy to improve the location update formula of the sparrow search algorithm to adjust the search area range. Finally, the improved algorithm proposed is applied to solve the 3D path planning problem of underwater robots. The simulation results show that the optimization performance of the potential field sparrow search algorithm (PFSSA) proposed in this paper is better than the particle swarm optimization algorithm (PSO) and the sparrow search algorithm (SSA), and it can quickly obtain an optimal and safe feasible path, which proves the effectiveness of the method, and it has application value in the field of path planning of underwater robot.

The application of logic circuit in catering industry is gradually increasing and diverse, one of the most important is to quickly and accurately control the cabinet door with each food material through logic circuit, so as to save the time of chefs to take food materials. Therefore, this paper takes 8 kinds of dishes and 10 kinds of ingredients of a sushi bar as the case and designs the logic circuit that can control the door of each sushi ingredients compartment through the truth table and the algebraic method of logic function. When inputting the serial number of each dish, the door of the compartment corresponding ingredients can be accurately opened. Meanwhile, Logisim software is used to draw the circuit diagram and verify the logic circuit design is true and effective.

The system is based on laser scanning technology to detect whether there is intrusion across the warning line or end of the train station; The trackside dynamic monitoring system can judge the train running direction, detect the train running speed, calculate the arrival time and transmit the data to the network control system; The control system can give different levels of audible and visual alarm according to the crossing or intrusion behavior and the arrival time of the train, and transmit the information to the staff for processing according to the situation. The designed platform intelligent monitoring system reduces the workload of operators, and improves the safety performance and accuracy.

Spacecraft thermal test temperature control systems are diverse and complex, and the control parameters corresponding to different temperature control systems are also different. This paper selects the identification model, determines the input signal and error signal and their difference, and calculates the system model parameters through the error criterion function. The identification method based on the least square method is adopted, and the error criterion function between the selected controlled object and its minimization mathematical model is used as the standard, and all the required data of the controlled system model is collected, and then the calculation is performed to estimate the temperature control system Parameters. Comparing the temperature data obtained by this identification method with the temperature data of the actual system shows that the identification accuracy is greater than 99%.

The classical vector control technology is usually adopted in doubly fed induction generator (DFIG). As the stator flux linkage cannot change suddenly, the grid voltage asymmetrical dips will generate transient DC flux linkage on stator side. At the same time, high electromotive force induced on the rotor side by accompanying negative sequence flux linkage, may exceed the maximum voltage of the rotor side converter. Hence the current is out of control due to the saturated converter. To solve the problem, a dual control is studied in the paper which can optimize the fluctuation of rotor current, stator current and electromagnetic torque during asymmetrical grid voltage dips. A simulation model of DFIG is built on Matlab/Simulink platform, and the two kinds of control strategies are compared and analyzed on rotor current, stator current and electromagnetic torque. The simulation results verify the effectiveness of dual control.

Traditional ATS (automatic test systems) are generally developed for instrumentation, which is less versatile and less reusable. Therefore, designing a reasonable, reliable, and efficient method of instrument and path allocation is a necessary part of improving the versatility of ATS (automatic test system). In order to solve the problem of channel selection and universality between UUT (Unit Under Test) and test instruments, firstly, UUT, instruments, test stations, etc. are described according to the IEEE1671 standard in the ATML (Automatic Test Markup Language) standard family; then, Through the matching of the UUT's demand signal and the instrument's capability signal, an instrument and path allocation algorithm is designed; finally, the algorithm is verified by an example of an airborne device, and the results show that the algorithm can effectively complete the instrument matching and allocate appropriate path to verify the feasibility and generality of the method.

Recently, detection algorithms for objects with arbitrary orientation have been greatly developed, but traditional object detection algorithms use horizontal boxes to label objects in images, which will lead to the problems such as loss of accuracy and excessive background information. In this paper, a rotation detection algorithm based on YOLOv5 is proposed, which can effectively detects the targets in the remote sensing image. It uses the CSL method to mark the oblique targets, converts the angle of the rotation boxes into a classification problem, and avoids the boundary problem. We also optimize the loss function to speed up the network convergence and adopt the NMS algorithm based on the rotation boxes, which effectively removes the redundant prediction frame. The experimental results demostrate that compared with the original YOLOv5 algorithm, the improved YOLOv5 algorithm can better locate rotation objects in remote sensing images and effectively prevent occlusion problems in dense target scenes. With a guaranteed inference speed of 24.1fps, Our method has a detection accuracy of 72.65% on dataset DOTA 1.0.

Aiming at the multipath effect existing in underwater wireless optical communication system, the small angle approximation method was used to simulate the multipath effect under different water quality and communication distance, and a new MDPCM modulation system was designed to suppress the multipath effect by changing the guaranteed time slot. The experimental results showed that the anti-multipath effect ability of the system could be improved and its bit error rate could be reduced by changing the guaranteed time slot in the MDPCM.

This paper, originally published on 2 December 2022, was withdrawn on 7 July 2023 per author request.

A relationship model between manufacturing data and complex equipment quality is studied and constructed. Combined with the data formed in each link of the whole life cycle of complex equipment, this paper studies and reveals the impact of manufacturing data on the quality of complex equipment. Based on the regulated intermediary theory model, the basic model of the regulation and intermediary relationship between product data richness, product quality inspection standard severity, product manufacturing risk and product quality is studied and established, and the action mechanism of product data richness on the quality of complex equipment is revealed, Through the case study, it is proved that the product manufacturing risk plays an intermediary role in the formation mechanism of the relationship between product data richness and product quality, and the severity of product quality inspection standards plays a regulatory role in the relationship between product data richness and product manufacturing risk, which provides a data-driven theoretical method for the quality control of complex equipment.

Aiming at the stability problem of unmanned surface vehicle (USV) heading system, an intelligent heading control strategy based on improved genetic algorithm is proposed considering the defects of poor robustness and low control accuracy of conventional PID control. First, the second-order nonlinear Nomoto response model of the USV is used to design the heading controller. Then, the genetic algorithm is used to optimize the heading controller parameters, and the adaptive crossover operator and variational operator are designed in the traditional genetic algorithm, and the improved genetic algorithm is applied to the USV intelligent heading control system. Finally, the optimization effect of the genetic algorithm on the heading controller before and after the improvement is compared and analyzed, and the simulation results show that the improved genetic algorithm effectively improves the response speed and control accuracy of the system.

Nowadays, embedded software has become the core part of weapons and equipment, with poor software quality being the bottleneck that constrains the development of weapons and equipment. Applying modularization technology to improve software quality and development efficiency by means of software re-use has become an important content of engineering research.

In the context of Peak Emission & Carbon Neutrality, focus on carbon dioxide emissions of BEV, make use of All-Round-STN (Traffic flow prediction network) and vehicle energy consumption model, predicted road state on predicted traffic flow rate, improve ACO as E-ACO, set energy consumption as the selection strategy, dynamic programming the path with the least energy consumption. Focus on range anxiety, combining shared road data and charging station data, build charging path selection algorithm. The development system provides optimal charging planning and intelligent travel service, which is based on the date of current power, the estimated energy consumption, the charging station layout, and the future road flow. Measuring emission reduction benefits based on public data, for carbon trading reference.

In the conventional mechanical bearing fault classification problem, deep learning has been widely used and has achieved very good results. However, in practical engineering, the lack of training data makes it difficult for the model to achieve the desired effect, and the model trained with the same data will also have the problem of insufficient generalization ability. Under the premise of small samples, this paper starts with transfer learning and data enhancement technology, uses quadratic interpolation, Mixup, sliding time window methods, and optimizes the model structure and hyperparameters to achieve the fault diagnosis of different bearings and working conditions a better classification effect.

Short rental apartments, Minshuku, also known as Homestay or B & B, which are usually scattered housings with frequent rental personnel replacement come along with security risks, and complex management problems. Intelligent door lock can solve the problems for the Minshuku owners by remote control. The research and development of door locks are changing with each passing day with more functions. However, in the scenario of Minshuku, factors that can affect the user experience of intelligent door lock haven’t been exhaustively investigated. Investigates with the Minshuku users were carried in this paper, factors affecting user experience were analyzed by APH and conclusions were drawn by Fuzzy Comprehensive Evaluation for intelligent door lock assessment.

The coil power supply is the core system of the EAST (Experimental Advanced Super-conducting Tokamak) device. In the field of nuclear fusion, the MDSplus database is often used to store experimental data and the built-in GUI tool jScope to display the experimental data. jScope can not meet the debugging needs. Based on the MDSplus database and the Python programming language, this paper develops the data analysis module of the power system of the EAST device, establishes the Python and MDSplus database interfaces, reads the data generated by the power system from the model tree, and calls the matplotlib package for waveform drawing. Visualize the power system data through Python, view the curve of voltage/current over time, obtain the difference, maximum, average, etc. of the voltage/current in any time period, and at the same time, you can read the relevant parameter values at any time point. The operation of the system plays a more precise and targeted debugging.

Energy storage in implantable electronic devices has become a popular area of research. The advances in this field improve the performance of implantable electrodes and the ability to integrate more functions inside such devices. Internal or external energy storage devices provide current implantable device power. This paper overviews two parts to the latest progress of in vivo energy storage devices and the self-powered implantable device. The energy storage devices in vivo include lithium batteries which have been widely used, highly biocompatible batteries that can reduce immune rejection in the human body, and the most emerging micro-supercapacitors. Four types of self-powered implantable devices are covered in this paper, including biofuel cells, biopotential devices, piezoelectric energy, and thermoelectric devices. Finally, the future prospect of this field is put forward. The current energy storage devices cannot fully achieve implantable devices with high biocompatibility long lifespan and are suitable for all organs simultaneously. However, the existing energy storage solutions can be selected according to different equipment needs and reasonably matched according to local conditions. The energy storage equipment can supply energy for implanted devices in the body safely, efficiently and with low risk as possible. This paper provides a reference for the selection of energy supply methods for implanted devices in the body in the future and provides directions for the research and development of new energy storage devices in the future.

In order to ensure the fire safety in the substation and improve the operation safety and reliability of the substation, this paper designs an intelligent window for substation fire protection based on single chip microcomputer. When the smoke sensor or combustible gas sensor detects that the indoor data is abnormal, it will automatically close or open the window at the same time, The design hardware uses STM32F407ZGT6 chip, MQ-2 smoke sensor, MQ-9 combustible gas sensor, ATK-VL53l0X laser ranging module, 28BYJ-48 stepping motor and buzzer. The program is designed on Keil uVision 5, and the debugged program is burned into the MCU for testing. The system shows good sensitivity and stability after a multitude of tests.

In this paper, W-SIC, a circuit with self-interference cancellation ability based on Wheatstone bridge is proposed to realize the physical layer absolutely safe communication in wired mode. The communication parties use a coaxial cable to transmit information to each other in different frequency ranges. They send information signals in their respective frequency ranges and release artificial noise signals in each other's frequency range. When the eavesdropper eavesdropping on the coaxial cable, the signals in both frequency ranges have been polluted by artificial noise, the eavesdropper can’t obtain effective signals. However, they can easily eliminate self-interference signals by using W-SIC, receive only the signal sent. Because the two signals are in different frequency ranges, the target information can be easily obtained by filtering, downconversion and other processing by the receiving. Besides W-SIC circuit model, a high-precision sliding rheostat is proposed to more accurately realize Bridge balance inside it, the optimal eavesdropping position of the eavesdropper is analyzed, the channel capacity of legitimate recipients, the formula of self-interference residual factor, the formula of system security capacity and the channel capacity of eavesdroppers are derived. The simulation results show the effectiveness and feasibility of the research method proposed in this paper.

Through the research on the temperature control system of the solid regenerative electric boiler, it is known that it has great limitations in terms of overshoot, control accuracy, and required adjustment time. In order to better improve it Based on the traditional PID algorithm, using advanced fuzzy control principles and strong learning ability of neural network, it is proposed to use fuzzy neural network PID controller to optimize temperature control to improve the accuracy and speed of temperature adjustment.This paper introduces the basic structure and principle of solid regenerative electric boiler, and proposes a fuzzy neural network PID algorithm for temperature control of solid regenerative electric boiler. To set up the conventional PID and the indistinct nervous network PID control model by means of the application of emulation. The summing-up is drawn: the indistinct nervous network PID controller has smaller overshoot and faster response in contrary to the conventional. In the case of interference, the anti-interference ability is more ideal.

Aiming at the problem that the arbitrary emission of wastewater from industrial enterprises, which has caused serious pollution of soil and groundwater in China, a set of intelligent control system for groundwater in-situ remediation equipment is designed in this paper. The system uses a programmable logic controller (PLC) as the main control unit, combined with pressure sensors, flow meters, ozone concentration detectors and other equipment, and uses injection well technology to achieve intelligent control of groundwater in-situ remediation equipment. In order to improve the performance of groundwater remediation, the injection module of the system can set a specific injection sequence and injection time for injection wells according to requirements. The system is of some significance for improving the automatic control level of groundwater treatment equipment.

In order to more quickly and accurately identify and repair the remote fault of hydropower station, aiming at the poor effect of remote fault management and control of hydropower station, a design method of hydropower station remote fault management and control system based on 5G inspection robot is proposed, the system hardware configuration is optimized, and the system software function and operation process are improved, In order to achieve the design goal of hydropower station remote fault management and control system. Finally, experiments show that the remote fault management and control system of hydropower station based on 5G inspection robot has high practicability and accuracy in the process of practical application, and fully meets the research requirements.

As a key link between information system and weapon, Shipboard Fire Control Radar urgently needs to improve its anti-jamming ability to ensure the operational effectiveness of radar in complex electromagnetic environment. To solve this problem, this paper proposes an anti-jamming strategy optimization method based on multiple decision network, which can reasonably allocate radar resources and make decisions quickly according to the electromagnetic spectrum distribution of environmental signals. The effectiveness of the proposed method is verified by field experiments.