最新刊期
卷 22 ， 期 8 ， 2023
- 摘要：The construction of the event logic graph for policy text can mine policy resource associations and share information， and assist policy formulation and publicity. This paper designs the policy title perceptual attention neural network to extract and classify policy events， based on the policy knowledge graph does data augmentation， formulates the extraction rules of the constituent relationship to realize the identification and extraction of the constituent policy event pairs， and combines the bottom-up thinking assessment policies relative importance of events，and form the event logic graph of policy text with the policy event as the node and the relative importance as the side. In the process of constructing the event logic graph of policy text experiment and the data augmentation method， the optimal baseline model of the proposed model increased by 0.79% and 0.16% in terms of recall and F-score indicators respectively. Finally， the construction and application processof a policy reasoning graph was demonstrated using bussiness environment policies as an example.关键词：event logic graph;policy text;knowledge graph;data augmentation;policy interpretation
- 90
- |
- 19
- |
- 0
发布时间：2023-08-25 - 摘要：Customer satisfaction is the customer's perceived evaluation of the degree of satisfaction with a product or service. Sentiment analysis technology can analyze customers' opinions， emotions and attitudes， and can evaluate customer satisfaction objectively and deeply. Taking online product reviews as the data source， the sentiment analysis technology based on machine learning is used to train and learn the sentiment polarity of each attribute review of the product， and the number of positive， negative and neutral reviews is obtained， which is processed by the intuitionistic fuzzy set method. The multi-attribute decision-making method based on intuitionistic fuzzy MARCOS is used to evaluate customer satisfaction， product attributes are used as evaluation indicators， and the review information in the form of intuitionistic fuzzy numbers is used as evaluation value. Finally， a new energy vehicle is taken as an example to conduct a customer satisfaction evaluation study to verify the feasibility and effectiveness of the proposed method.关键词：customer satisfaction evaluation;online review;sentiment analysis;MARCOS
- 41
- |
- 11
- |
- 0
发布时间：2023-08-25 - 摘要：Accurately predicting the identity of individuals involved in terrorism is of great significance for the anti-terrorism work of public security organs. To this end， in order to improve the efficiency of data processing and prediction in the mining of terrorism related intelligence， we comprehensively consider the five elements of public security information and establish a label system for terrorism related personnel， using labels to accurately describe the data characteristics of terrorism related personnel. At the same time， the evidence weight and information value are introduced to screen strongly related indicator variables， and the evidence weight logistic regression model is established to predict the identity of terrorist related persons， so as to solve the trap and over fitting problems of dummy variables. The experiment shows that the prediction effect of the proposed model on terrorist related personnel is better than that of logistic regression and other prediction models.关键词：terrorist-related personnel;identity prediction;labelling system;logistic regression;weight of evidence;information value
- 43
- |
- 7
- |
- 0
发布时间：2023-08-25 - 摘要：PM2.5 has a significant impact on air quality and poses a greater threat to human health compared to coarser atmospheric particles. To facilitate accurate prediction of PM2.5 concentration levels， PM2.5 concentration level prediction method based on multi algorithm fusion and humanistic features is proposed. Firstly， meteorological data from Lanzhou City from 2018 to 2021， the temporal variation of PM2.5 concentration at predicted stations， and the spatiotemporal correlation of surrounding stations were selected as characteristic factors； Then， the strong correlation degree of meteorological indicators and the correlation degree of various meteorological factors and pollutants were analyzed using grey correlation analysis and Pearson coefficient， respectively， and expanded into multiple features （humanistic features） using the time dimension； Next， based on multi-layer perceptron， random forest， LSTM and other algorithms， a model is established to predict PM2.5 level； Finally， based on the multi feature fusion algorithm， the final prediction model is constructed using the prediction results of each model. The experiment shows that the accuracy of the proposed method in predicting PM2.5 level reaches 89%， and the accuracy and F-value reach 88%， which is of great significance for predicting air quality and formulating relevant preventive measures.关键词：PM2.5;multi algorithm fusion;meteorological factors;humanistic feature
- 33
- |
- 7
- |
- 0
发布时间：2023-08-25 - 摘要：In recent years， helmet wearing detection has been widely studied in the field of deep learning. In order to solve the problems of low accuracy， weak anti-interference ability， and poor portability of existing helmet detection algorithms， an improved YOLOv5 based object detection algorithm is proposed. Firstly， the residual module is introduced to connect deep feature information with shallow feature information， enhancing information exchange between networks； Then， combined with the coordinate attention mechanism （CA）， the model's feature extraction ability is improved without increasing computational complexity； Finally， optimize the loss function curve to improve the regression effect of the model. The experiment shows that the proposed model improves the detection accuracy by 3.6% compared to the traditional YOLOv5 method， and can quickly and accurately detect safety helmets， making it more suitable for solving practical engineering problems.关键词：helmet detection;YOLOv5;coordinate attention mechanism;ResNet
- 52
- |
- 10
- |
- 0
发布时间：2023-08-25 - 摘要：Aiming at the problem that the traditional drug recommendation algorithm ignores the semantic information of the drug itself， which leads to the low recommendation accuracy， by constructing a drug knowledge graph and introducing the semantic information between drugs as an important recommendation basis， a TransR-CF drug recommendation integrating knowledge graph and collaborative filtering is proposed. algorithm. By using the knowledge representation learning translation model TransR， the drug knowledge graph is mapped to a low-dimensional continuous vector space to obtain drug entities and their relationships， thereby calculating the semantic similarity between drugs. The resulting behavior similarity is linearly fused for recommendation. Based on the traditional collaborative filtering drug recommendation algorithm， the algorithm also considers the semantic information between drugs， which reduces the impact of data sparsity and cold start on the recommendation results to a certain extent， making the recommendation results more reasonable and credible. The experimental results on the public data set of sales data of an online pharmacy in the Hejing community show that the algorithm has improved to a certain extent compared with the benchmark algorithm in the three indicators of precision， recall and F-measure.关键词：drug recommendation;collaborative filtering;knowledge graph;knowledge representation learning
- 21
- |
- 4
- |
- 0
发布时间：2023-08-25 - 摘要：The difficulty of dealing with pixel loss in semantic segmentation through deep learning is that the upsampling process cannot fully restore pixels， and most existing network models pursue high prediction accuracy， resulting in complex structures and reduced prediction efficiency， making it difficult to meet real-time requirements. A real-time multi-layer fusion attention mechanism semantic segmentation network model is proposed to address the above issues. This model accurately restores pixels from two aspects， achieving better real-time performance while ensuring high processing accuracy. Firstly， in the U-shaped network horizontal connection， an efficient horizontal connection module is constructed by combining attention mechanism with deep and shallow features. Its advantage lies in the attention mechanism focusing on more comprehensive contextual features to provide as much pixel information as possible for subsequent upsampling processes； Then， considering that the feature points generated by the deep network have stronger localization ability， the model uses deep features for pixel localization correction during the upsampling process， making pixel restoration more accurate and better solving the problem of edge smoothing in prediction results. Adopting a lightweight model as the basic model， using 1×1 convolution for dimensionality reduction in multiple locations during the fusion process. The experimental results show that when the input image size of the VOC and Cityscapes datasets is 512×1 024， the model can maintain a processing speed of 64 frames/s， with an average intersection to parallel ratio of up to 75.3%.关键词：computer vision;multi-layer fusion;pixel restoration;edge smoothing
- 34
- |
- 4
- |
- 0
发布时间：2023-08-25 - 摘要：In order to analyze the emotional polarity of peer-reviewed texts and balance the subjectivity of manual scoring， a BERT-MSCNN sentiment classification model combining multi-scale convolutional neural network and channel attention mechanism is proposed. Firstly， the crawler is used to crawl the public paper review data of ICLR conference from the OpenReview platform and translate it into Chinese to construct a peer review data set. Secondly， BERT pre-training is used to complete the word vectorization of the input statement， and local semantic features are extracted by multi-scale convolutional neural network. Finally， the extrusion-excitation network SENet is introduced to calibrate the feature channels， so as to enhance the feature extraction and improve the classification effect of the model. The experimental results show that the proposed model can effectively classify the sentiment of peer review data， and its accuracy can reach 78.12%. Compared with the existing text classification methods based on CNN and LSTM， it has higher recognition rate.关键词：peer review;sentiment classification;deep learning;convolutional network;channel attention
- 12
- |
- 3
- |
- 0
发布时间：2023-08-25 - 摘要：Most existing multi-target tracking algorithms only consider the features themselves and seldom consider other features within frames and the interaction of features between frames， two methods are proposed to make the features more discriminative in order to achieve the interaction between features， enhance the feature representation， obtain a more reliable similarity matrix， and improve the accuracy of target association. The first method is to introduce graph neural network in feature processing， and build appearance graph neural network and motion graph neural network respectively， the features in the two network models fuse and complement each other to enhance the expression of target features and improve the feature discriminative property； the second method is to introduce update mechanism in graph neural network. Previously， graph neural networks are mostly static graphs， which cannot be dynamically updated with applications. To this end， a strategy is proposed to make the edge module， point module and global variable module in the graph neural network dynamically updated， where the global variable can capture the global relationship of features in the graph network to better accomplish target matching. Theoretical analysis and tracking algorithm simulation results show that the accuracy metrics （MOTA） of the proposed multi-target tracking algorithm are 49.5% and 53.9% in two benchmark test sets， MOT16 and MOT17. Meanwhile， the performance of the three update modules of the graph neural network is compared by ablation experiments， and it is found that the graph neural network with the complete structure module improves the MOTA index by 1.9% and 0.2% than the module without the point feature aggregation module and the module without the global variables， respectively， indicating that the multi-target tracking algorithm based on the graph neural network has better performance in the actual tracking task.关键词：computer vision;graph neural network;multi-object tracking;data association;appearance model;motion model
- 14
- |
- 7
- |
- 0
发布时间：2023-08-25 - 摘要：Sentiment analysis of consumer reviews is of great significance to the study of user needs. Currently， consumer comment sentiment analysis typically uses word2vec or Glove models to generate static word vectors， which fail to effectively integrate contextual context and solve the problem of polysemy. Additionally， a single neural network model has a relatively simple structure and cannot effectively extract important features when faced with complex comment texts. Therefore， a model based on BERT and attention mechanism is proposed. Firstly， using BERT for word vectorization， a dynamic word vector containing contextual semantic information is generated； Then， the attention mechanism is used to capture the relationship between the global and local， and the extracted vectors are input into the Fasttext module for final sentiment classification. The experimental results show that the accuracy of the proposed model reaches 95.16% on the Tan Songbo Hotel Review dataset， and 93.65% on the IMDB dataset. Compared with the single model， the performance of the proposed model is significantly improved， which has certain practical significance.关键词：sentiment analysis;BERT;attention mechanism;FastText
- 21
- |
- 4
- |
- 0
发布时间：2023-08-25 - 摘要：In response to the shortcomings of traditional table lookup and convolution methods in acceleration and deceleration control， such as long running time， discontinuous acceleration， and acceleration， a fourth order S-shaped acceleration and deceleration continuous integration control algorithm based on field programmable gate array is proposed on the basis of analyzing the principles of trapezoidal and S-shaped acceleration and deceleration algorithms. The algorithm uses Verilog hardware description language to write motor control program， and uses Modelsim software to simulate acceleration and deceleration functions.The experimental results show that compared with the table lookup method and convolutional method for acceleration and deceleration control， this algorithm can shorten processing time， reduce motor out of step， improve work efficiency， and is of great significance for optimizing motion control systems.关键词：motion control;four S-shaped acceleration and deceleration;continuous integral;FPGA
- 19
- |
- 2
- |
- 0
发布时间：2023-08-25 - 摘要：Aiming at the problems of excessive fluctuation of the number of clusters， uneven distribution of cluster heads， and unbalanced energy consumption in the LEACH protocol in wireless sensor networks， A WSN clustering routing algorithm GAKFCM-LEACH （genetic algorithm and kernel fuzzy C-Means based LEACH）based on GA-KFCM clustering algorithm and LEACH protocol is proposed. In the clustering stage， the base station uses the GA-KFCM algorithm to cluster the nodes in the working area to ensure that the number of clusters is stable and the size of the clusters is relatively equal. In the cluster head election stage， the remaining energy of the node， the distance of the base station， and other factors are considered， and the weight factor is dynamically configured. In the data transmission stage， the current cluster head energy consumption in different transmission paths is estimated and used as an important indicator to determine the next hop cluster head， so that the multi-hop transmission path can be optimized. The simulation results show that compared with the LEACH protocol and its improved algorithm， the algorithm makes the clustering more stable， the network lifetime is 108% longer than that of LEACH， and 20% and 15% longer than that of KMEANS-LEACH and FCM-LEACH，respectively.关键词：GA-KFCM clustering;wireless sensor network;LEACH;routing;multi-hop
- 9
- |
- 1
- |
- 0
发布时间：2023-08-25 - 摘要：Reinforcement learning has been widely used in various industries. The large amount of exploration required for reinforcement learning， however， can have serious consequences in many contexts， and the industry has proposed safe reinforcement learning to address this. The utility function， borrowed from economics， is one of the commonly used techniques， but its application in search algorithms has not been thoroughly investigated. How to use search algorithms to migrate an agent trained in a risk-neutral context to a risk-sensitive context has to be studied. To address this， it is proposed to improve the Expectimax search algorithm by integrating the utility function trick commonly used in safe reinforcement learning with the vanilla algorithm. Taking the 2048 game as an example， based on the weights trained in the risk-neutral context using the latest NTuple method in 2022， different hyperparameters were tested， and applying the sigmoid function as the utility function eventually led to higher mean scores： the improved algorithm achieved an increase of 14 053.35 points for 2-level search and 6 522.9 points for 3-level search compared to the original method， demonstrating the effectiveness of the proposed improvement.关键词：random games;game 2048;expectimax search;safe reinforcement learning;utility functions
- 12
- |
- 5
- |
- 0
发布时间：2023-08-25 - 摘要：The random division based isolation forest （iForest） algorithm does not consider the probability of outliers in subsamples. An isolated forest algorithm based on range（r-iForest） which utilizes range to select sample subsets in the random sub-sampling process is proposed. The approach results in sample subsets with higher probability of outliers. During the construction of the isolation tree， the topology shape of the tree is controlled by the proportion of the child node and its immediate parent node in terms of sample size， so as to avoid an isolation tree with poor performance is generated. The results of comparison with eight outlier detection algorithms on seven public datasets from ODDS and the KDD CUP 99 datasets show that the accuracy of r-iForest algorithm is 2 to 40 percentage points higher than other algorithms， and the time consumption of r-iForest algorithm is about 15% lower than that of iForest algorithm.关键词：random subsampling;outlier detection;isolated forest algorithm;range
- 10
- |
- 10
- |
- 0
发布时间：2023-08-25 - 摘要：Coal mine safety production has always been concerned. How to ensure miners' personal safety and determine the fuzzy position of hidden danger has always been the research focus in the field of coal mine safety production. Aiming at how to determine the dangerous position in the complicated underground environment， this paper puts forward a method of distinguishing the fuzzy position of underground dangerous situation. The collected miners' voice data is converted into text data at the mobile terminal， and fuzzy data is completed for this data， and the data set is completed by priority strategy； Finally， combined with the status of miners， the dangerous position is determined by centroid algorithm.关键词：coal mine safety;hidden danger;position discrimination;text data;data processing
- 5
- |
- 1
- |
- 0
发布时间：2023-08-25 - 摘要：When traditional genetic algorithm （GA） is used to solve traveling salesman problem （TSP）， the global search scope is limited， and it is easy to fall into premature phenomenon and slow Rate of convergence due to the lack of population diversity and local search ability in the later stage. Based on the population inversion genetic algorithm （EROGA）， which combines the idea of reverse learning （OBL）， the Metropolis principle of Simulated annealing algorithm （SA） and the idea of real elite learning， an improved genetic algorithm OBLGSAA is proposed. First， reverse learning is adopted in the process of generating initial population to improve the accuracy and Rate of convergence of the optimal solution； Then， the Metropolis criterion is used to improve the crossover and mutation operators to enhance the local search ability of the algorithm； Finally， the concept of real-life elite learning is introduced to further enhance GA's local search ability through a greedy rotation learning mechanism. The simulation results on various traveling salesman datasets show that OBLGSAA can effectively improve the diversity of GA population， making the algorithm less prone to premature convergence， and outperforming the original EROGA in terms of convergence performance and solution accuracy， verifying the effectiveness and superiority of the algorithm.关键词：opposition-based learning;Metropolis criterion;genetic algorithm;traveling salesman problem
- 30
- |
- 5
- |
- 0
发布时间：2023-08-25 - 摘要：With the rapid development of economy， the flight flow of regional Airline hub is increasing. If the short-term flow is too large， congestion is likely to occur. The control department will take flight flow control measures to solve the congestion problem， resulting in a large number of flight delays and increasing the workload of controllers. To address the above issues， a high-precision entrance and exit flow prediction model based on Prophet is proposed. This model fully considers seasonality， trendiness， and vacation characteristics in the time series， and adopts different fitting functions. It has good predictive performance on datasets with strong periodicity. Select the high-precision takeoff and landing message statistical data between 2019-07-27 and 2019-09-25 of Shuangliu Airport， and construct the inbound and outbound flow data set after Data cleansing and Z-SCORE standardization. According to the characteristics of the actual traffic curve， a daily fixed time period was selected as the change point to adjust the Prophet prediction model， and visual parameter adjustment was performed using images. The obtained results were compared with the results of models based on ARIMA and LSTM. It was found that the proposed model had a 17.9% improvement in MAE， 52.0% improvement in RMSE， and 50.82% improvement in MAPE compared to the LSTM algorithm with the closest performance. At the same time， in order to address the limitations of a single feature in flow prediction， multiple meteorological feature data were introduced， and feature transformations were carried out based on actual flight conditions. Combined with the LightGBM algorithm， the Prophet prediction results were optimized and corrected. The final result was further improved compared to the original Prophet， and correctly reflected the impact of climate change on flight flow.关键词：Prophet;LightGBM;METAR;short-term airport flow prediction;time series
- 8
- |
- 1
- |
- 0
发布时间：2023-08-25 - 摘要：In the field of software test automation， the method of generating test cases about path coverage has always been a research hotspot. The traditional genetic algorithm， which constructs a new fitness function， requires the fitness function to be recalculated in each iteration process， which increases the amount of calculation. Therefore， a path test case generation method based on branch evolution theory is proposed. First calculate the node complexity of all nodes in the basic path set， consider the branch complexity from the node to the branch and improve the fitness function， and sort the target path according to the node complexity； when using the genetic algorithm to generate test cases covering the target path， after the subpopulation covers the current target path， it continues to cover other paths according to the distance of the target path. The experimental results show that the method has a significant improvement in evaluation times and running time compared with other existing algorithms.关键词：test case generation;path coverage;genetic algorithm;branch evolution;target path distance
- 10
- |
- 2
- |
- 0
发布时间：2023-08-25 - 摘要：With the widespread application of commercial SMS services， excessive commercial marketing has brought troubles to mobile users and also affected the normal regulatory order of the industry. The business process of commercial SMS services usually generates corresponding communication record information （CDR）. However， under the existing commercial SMS service model， the lack of transparency in the identity of business entities will make it difficult to trace CDR. When dealing with user complaints or violations， industry regulatory authorities need to spend a long time verifying the effectiveness of SMS CDR. To this end， a blockchain based commercial short message supervision method is proposed， which introduces blockchain technology into traditional supervision models and combines smart contracts to achieve trustworthy and automated management of commercial short message CDRs. This enhances the transparency of the identity of commercial short message service entities， provides technical support for CDR traceability， and ensures the authenticity and trustworthiness of distributed ledger CDRs. This not only reduces the time cost of the supervision process， but also improves regulatory efficiency， hoping to provide new solutions for the regulatory model of commercial short message services.关键词：commercial SMS supervision;blockchain;distributed ledger;smart contract;traceability
- 17
- |
- 3
- |
- 0
发布时间：2023-08-25 - 摘要：The ecological protection and high-quality development of the Yellow River Basin have received widespread attention as a major national development strategy. As an important soil and water conservation engineering measure in the Loess Plateau region， warping check-dams have played a very important role in controlling channel erosion and retaining sediment. However， after many years of operation， most silt check-dams have encountered problems such as large sediment deposition and aging of the dam body. Once the hydrological and meteorological conditions change， it will seriously affect the safety of the dam system and watershed. In order to solve the problems of low informatization level of warping check-dam management and imperfect flood control early warning mechanism， a simplified early warning model for warping dam in the Yellow River basin of Inner Mongolia was proposed based on comparative analysis and expert discussion， and a flood control early warning system that can realize visualization of warping dam management information and early warning Information visualization was designed. The feasibility of the model was verified through experiments based on real data， providing technical support for the safe flood control of the silt check-dam project and comprehensively improving the informationization level of silt check-dam management.关键词：check-dam;early warning of flood control;analytic hierarchy process;RUSLE
- 12
- |
- 3
- |
- 0
发布时间：2023-08-25 -
Design and Implementation of Precision Control Platform for Installment Agriculture Based on the IoT AI导读
摘要：Aiming to achieve intelligent and precise control of facility agriculture， reduce labor costs and minimize crop pests and diseases， the design and implementation of facility agriculture greenhouse scenarios in combination with IoT technology were made to create a facility agriculture precision control platform based on IoT. The platform architecture is described through a highly abstract six-domain iot model architecture. Communication between SCD and ASD is achieved through building MQTT message cloud services and using DTU device protocols. The platform has five modules including environment perception module， greenhouse pest and disease identification module， equipment control module， through which environmental data such as soil moisture and temperature， relevant nutrient element content， air light intensity within the greenhouse can be collected， and the crop pests and diseases will be recognized. Ultimately， precise control of facilities can be achieved while assisting farmers to scientifically manage their mobile devices.关键词：installment agriculture;six-domain model;Internet of Things;precision control;greenhouse- 14
- |
- 6
- |
- 0
发布时间：2023-08-25 - 摘要：In view of the problems that the existing online evaluation system is used in the practical teaching of programming courses， such as low students' learning investment， single examination process and lack of dynamic expansion and contraction. Therefore， the existing online evaluation system is redesigned and implemented， and the comprehensive evaluation mode combining open-book training with closed-book testing is adopted， and the micro-service and Docker technology are applied to restructure the system. According to the analysis of stress test， the new system can cope with high concurrency， and the comprehensive evaluation mode can improve students' learning input， which has reached the system design goal.关键词：programming;online judge system;closed-book testing;micro-service architecture;Docker
- 43
- |
- 12
- |
- 0
发布时间：2023-08-25 - 摘要：In order to improve the ability of urban emergency management in early warning， disposal and recovery under the Big data environment， the research literature on the application of intelligence， Big data and text semantics to urban emergency management in the past five years was selected to review with HowNet as the retrieval platform， and the main ideas， method models， shortcomings， challenges and opportunities of the application of Big data technology in urban emergency management were summarized. The use of Big data technology to integrate information and improve government functions is conducive to improving urban emergency management capabilities， so as to ensure the safe and stable development of the city.关键词：big data;urban emergency management;smart city
- 14
- |
- 2
- |
- 0
发布时间：2023-08-25 - 摘要：Aiming at the demand of high-speed data transmission for laser scanning confocal microscope， a large data caching and high-speed transmission system based on DDR3 and gigabit Ethernet is designed. The system uses Artix-7 series FPGA chip XC7A35T as the main control chip， and designs the control scheme of the whole system on Xilinx FPGA development platform Vivado with Verilog language. The cache module uses Micron's DDR3 chip MT41J128M16HA， combined with the DDR controller IP core and FIFO provided by Xilinx， to cache a large amount of upstream data. At the same time， Realtek's Ethernet chip RTL8211 is used to realize Ethernet frame grouping and unpacking based on UDP and IP protocols. It receives instructions from the host computer or sends data to the host computer at a rate of 1 000 Mbps through RJ45 Ethernet interface. The test results show that the scheme can store data in DDR3 at high speed and transmit it to the host computer efficiently in real time， and the real-time transmission rate reaches 118.75 Mbyte/s. The system has high stability， low bit error rate， high circuit integration and precise control timing， which is very suitable for the data transmission system of laser scanning confocal imaging equipment.关键词：confocal microscopy;FPGA;DDR3 SDRAM;gigabit Ethernet;Verilog;caching and transport
- 11
- |
- 1
- |
- 0
发布时间：2023-08-25 - 摘要：Smart contracts are an important component of blockchain， characterized by decentralization， traceability， automated execution， and non tampering. They can effectively meet the security requirements of drone clusters at both the theoretical and technical levels. Taking the application of smart contracts in flight data management， autonomous collaboration， security maintenance， and security authentication in unmanned aerial vehicle clusters as an example， this paper focuses on analyzing the potential security vulnerabilities of smart contracts in the application process， including integer overflow， timestamp， reentry， transaction sequence dependency， and transaction authorization. Based on this research， a hybrid neural network vulnerability detection model based on attention mechanism is proposed for potential security vulnerabilities in smart contracts. Experiments have shown that compared to the current popular smart contract vulnerability detection technology， the proposed vulnerability detection model ACBSC has better detection performance， and has better accuracy and precision in detecting vulnerabilities. The research results have certain practical significance for improving the safety factor of drone cluster construction， and also provide reference for unmanned construction and development.关键词：smart contract;drone swarm;attention mechanism;hybrid neural network;vulnerability detection
- 10
- |
- 2
- |
- 0
发布时间：2023-08-25 - 摘要：While visiting some special web pages， such as hospitals， schools， banks， etc.， web traces， which record a series of requests and responses generated when visiting a web page， may reveal personal privacy. Traffic analysis attacks can easily identify the page a user is visiting， even if the transmitted traffic is encrypted， and infer the user's privacy. This paper proposes a new method to defend against traffic analysis attacks by padding the size of web objects on the server side without changing the rendering effect， introduces the concept of rendering equivalent and makes the object sequence size and order the same for both pages by adding NULL objects and padding objects. Following the idea of K-anonymity， in a set of web pages with privacy， we use a similar website fingerprinting strategy to obfuscate classifiers to identify web pages， and achieve good results. What's more， the relevant statistical analysis of the data is still available after applying our strategy. Finally， the proof-of-concept experiments is conducted to confirm the feasibility of our strategy.关键词：traffic analysis;website fingerprinting;similar traffic;classifier;network security
- 12
- |
- 3
- |
- 0
发布时间：2023-08-25 - 摘要：Most methods use auto-encoder or generative adversarial networks to hide secret information such as image or audio in the cover image. Steganography can ideally generate stego image which is similar to the cover image， so that others cannot recognize the secret information. Compared with traditional methods， the methods based on deep learning are not ideal. Therefore， a steganography model （W-AE） is proposed in this paper. The encoder of W-AE will fuse cover and secret image， both which need to be processed by discrete wavelet transform. The decoder network will recover the secret image. Experiments show our method can generate high-quality stego image whose average PSNR and SSIM between stego and cover image are 33.9 and 0.91. The model has better security than existing methods， and can cheat the steganalysis YeNet and ATS in unknown weights. Besides， the influence of wavelet transform on the embedding position of secret information is analyzed to increase the interpretability of the model.关键词：steganography;wavelet transform;auto-encoder;information-hiding
- 7
- |
- 1
- |
- 0
发布时间：2023-08-25 - 摘要：Aimed at the problems of low convergence accuracy and too many parameters of human mental search algorithm and pulse coupled neural network， which lead to poor quality of fused images， a PCNN image fusion method based on parameter adaptation is proposed. First， an algorithm that introduces uniform grouping， local search and the move count operator into the human mental search algorithm is proposed to overcome the problems of poor search ability and slow convergence of HMS； Then， the PCNN's αθ，β，αL parameters are optimized by the improved algorithm to enhance its feature extraction performance； Finally， the complete fused image is obtained by fusing the ToF and RGB images with the PCNN after parameter optimization. Twenty test functions were selected for simulation experiments of the algorithm， and the experimental results showed that the proposed algorithm can effectively speed up the convergence speed and improve the accuracy of the optimization search. The effectiveness of the method was verified by selecting orchard images under different temporal lighting， and the average index of the proposed method was improved by 19.99%， 30.74%， 21.14%， 17.6%， and 8.93% compared to DWT， TIF， CNN， PCNN， and DWT_PCNN， respectively. The method can significantly improve the quality of fused images.关键词：human mental search algorithm;pulse coupled neural network;image fusion
- 12
- |
- 2
- |
- 0
发布时间：2023-08-25 - 摘要：Existing face image inpainting methods， which synthesize face images under the conditions of large scale image deficiency or high resolution that will have inconsistent image texture structure and inconsistent contextual semantic information， and cannot accurately synthesize clear image structures， such as eyes and eyebrows， etc. To solve this problem， This paper proposed a Rough-Local-Global Networks（RLGNet） face image inpainting method based on a Normalization-based Attention Module（NAM）， which can accelerate the convergence speed of the model and reduce the computational cost. Among them， the coarse inpainting network performs the initial repair of the residual image and then the local inpainting network performs the detailed repair of the residual image locally； the global refinement inpainting network based on the normalized attention mechanism performs the repair of the residual image as a whole to increase the semantic coherence and the coordination of the texture structure of the image. The method proposed in this paper is tested on the CelebA-HQ dataset. The results show that the PSNR reaches 30.35 dB and the SSIM value reaches 0.926 9 and FID is 2.55 at the mask ratio of 20%~30%， which can synthesize face images with a natural transition of adjacent pixels and a reasonable texture structure.关键词：face image inpainting;normalized attention module;three-stage inpainting network;activation function
- 9
- |
- 4
- |
- 0
发布时间：2023-08-25 - 摘要：A liver cancer image segmentation method based on UNet++combined with residual network and attention mechanism （RA-UNet++） is proposed to address the issues of insufficient contextual information connections and tumor morphology differences in liver cancer image segmentation， resulting in difficulty in model feature extraction and low similarity coefficients. Residual connections can provide more low-level features for the model to assist in accurately locating tumors； By assigning attention weights to feature map channels through attention mechanisms， the model focuses on special areas in the image and extracts more discriminative features. The experiment on the LiTS-ISBI2017 liver tumor dataset showed that the proposed model achieved volume overlap error， volume difference， and similarity coefficient of 3.01%， 0.25%， and 98.65%， respectively， confirming the effectiveness of the method and providing reference and reference for segmenting liver cancer images.关键词：UNet++;residual network;attention mechanism;image segmentation
- 22
- |
- 4
- |
- 0
发布时间：2023-08-25 - 摘要：In order to solve the problem of false-positive blood vessels and missing microvessels in the segmentation of fundus images by existing methods， a generation adversarial network suitable for retinal blood vessel segmentation is proposed. The generator of the network uses U-Net structure to generate high-quality false images， and the discriminator uses fully connected convolutional neural network to improve the discrimination ability. The proposed generation countermeasure network model is tested on the datasets DRIVE and STARE. The experimental results show that the accuracy and sensitivity of the proposed segmentation algorithm are significantly improved， which are 2.84% and 1.44%， 5.89% and 7.96% higher than the original U-Net algorithm respectively， and 1.91% and 1.08%， 4.94% and 6.88% higher than the original OR-GAN algorithm respectively. The segmentation of retinal vessels also has better performance than other existing methods.关键词：medical image segmentation;retinal vascular image;generate adversarial network;U-Net algorithm;fully connected convolutional neural network
- 15
- |
- 3
- |
- 0
发布时间：2023-08-25 - 摘要：Aiming at the problem that the traditional Canny algorithm is sensitive to salt-and-pepper noise， only vertical and horizontal directions are considered when calculating the gradient amplitude， and only the gradient amplitude in the gradient direction is used in the process of non maximum suppression， which leads to inaccurate gradient calculation and blurred image edges after filtering， an improved Canny edge detection algorithm is proposed. First， the hybrid filter combining adaptive median filter and guided filter can not only effectively remove salt-and-pepper noise in the image， but also better preserve the image edges； Then， use the Sobel operator template in four directions to calculate the image gradient； Finally， in the process of non maximum suppression， the adaptive linear interpolation method is used to improve the edge detection accuracy. Simulation results show that the proposed algorithm can not only effectively filter salt-and-pepper noise， but also detect more edge details than the traditional Canny algorithm.关键词：image edge detection;Canny algorithm;hybrid filtering;sobel operator in 4 directions;adaptive linear interpolation
- 17
- |
- 1
- |
- 0
发布时间：2023-08-25 - 摘要：Project based learning is an effective way to implement moral cultivation， cultivate core competencies and answer "how to cultivate people". However， at present， the integration of project-based learning and curriculum teaching is relatively shallow， more attention is paid to the general research of project-based learning， and no effective method is pointed out to deeply integrate project-based learning and discipline teaching. Therefore， this research adopts the method of systematic literature review to systematically analyze the relevant literature of project-based learning in the field of international educational technology through literature measurement， literature research and specific cases. The research finds that project-based learning is an effective learning mode based on constructivism， pragmatism and situational learning theory， which is problem driven to promote the development of students' core literacy； The basic elements of project-based learning include real situations， driving problems， collaborative activities， reflective evaluation and project results. In addition， the research on project-based learning mostly adopts empirical research methods and mainly focuses on the field of higher education. These studies show that project-based learning has a positive effect on improving learners' achievements， performance， teamwork， self-awareness， emotion， behavior， etc. Based on this， the research attempts to put forward suggestions on the deep integration of core competencies and project-based learning in China： taking the cultivation of core competencies as the core goal of project-based learning， building a project-based learning design model based on core competencies， building an evidence oriented project chemistry evaluation system， and introducing achievements in other fields to improve project-based learning based on core competencies.关键词：educational technology;project-based learning;literature review;core competencies;learning mode
- 7
- |
- 1
- |
- 0
发布时间：2023-08-25 - 摘要：Artificial intelligence and education have been deeply integrated. The foundation of rural revitalization is the revitalization of rural education. Artificial intelligence can provide support for rural education revitalization because of its characteristics of supporting personalized learning. Based on SWOT-CLPV model analysis， this study analyzes the support of artificial intelligence for rural education revitalization.The research results demonstrate the internal advantages and disadvantages， external opportunities and threats of the application of artificial intelligence in rural education revitalization. Further， based on SWOT-CLPV model， the paper analyzes the leverage， inhibition， vulnerability and problems of artificial intelligence supporting rural education revitalization. On the basis of the above analysis， this paper puts forward some suggestions and countermeasures for AI to support rural education revitalization.关键词：artificial intelligence;rural education revitalization;SWOT-CLPV model
- 7
- |
- 4
- |
- 0
发布时间：2023-08-25 - 摘要：In the intelligence era， emerging information technologies are constantly innovating and advancing. The promotion of education modernization has put forward new demands on the modern information technology teaching in colleges and universities， which traditionally emphasized knowledge teaching but neglected literacy training. Aiming at cultivating information core literacy， this paper outlines the basic connotations of information core literacy in modern information technology teaching in colleges and universities. From the four dimensions of cross-disciplinary theoretical knowledge， cross-temporal experience learning， cross-media migration skills， and cross-domain technology application， this paper explores the cultivation path of multi-factor information security awareness， multi-stage information computing thinking， multi-level information outreach ability， and multi-environmental information ethical responsibility based on "attack and defense combination， virtual and real communication， internal and external connection， as well as moral and intellectual synergy". Moreover， the practical exploration of information core literacy cultivation path is carried out based on the course on modern information technology and artificial intelligence. As a result， the cultivation framework of information core literacy in modern information technology teaching is constructed to meet the development requirements of education modernization.关键词：information core literacy;modern information technology teaching;education modernization;information security awareness;information computing thinking;information outreach ability;information ethical responsibility
- 10
- |
- 2
- |
- 0
发布时间：2023-08-25 -
Education Informatization 2.0 Promoting the Digital Transformation of Education in Southern Xinjiang AI导读
摘要：Under the background of education informatization 2.0 era， new requirements are put forward for the development of education. The vigorous development of digital education resources is of great significance to the construction of a powerful country in education and human resources. Educational cooperation and dialogue is the core of constructivism theory，its purpose is to improve the originally weak quality of education and teaching， make the utilization of digital education resources really take root in each class， and meet the transformation from group needs to individual needs. Through the exploration of the restrictive factors of the development of rural digital education resources in southern Xinjiang in the era of educational informatization 1.0， this paper defines the connotation and characteristics of digital education resources in the stage of educational informatization 2.0， outlines the mode of digital education resources promoting the collaborative sharing of educational resources in southern Xinjiang， and tries to put forward the transformation path and strategy， in order to provide reference for subsequent research.关键词：educational information 2.0;digital education resources;southern Xinjiang region;digital transformation- 9
- |
- 2
- |
- 0
发布时间：2023-08-25
- Technical support is provided by Beijing Founder electronics co., LTD 鄂ICP备12016542号-70
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0