最新刊期
卷 22 ， 期 10 ， 2023
- 摘要：A scalable simulation workflow and a simulation module library are proposed for discrete and continuous system simulations in the automotive electronic control unit domain. Firstly， corresponding calculation methods are defined for commonly used basic simulation modules in the domain， along with their attributes and behaviors， establishing a basic simulation module library. Then， by preprocessing and topological sorting of the model graph， a module calculation sequence table is obtained. The simulator uses this sequence table as a reference to calculate the states of various basic simulation modules within a time step. Finally， a fourth-order Runge-Kutta solver is provided to support continuous event simulation. By comparing with the existing commercial software Simulink in typical cases from discrete， continuous， and hybrid domains， the accuracy and usability of this software are demonstrated.关键词：simulation software;discrete event simulation;continuous event simulation;simulator;solver
- 92
- |
- 22
- |
- 0
发布时间：2023-10-24 - 摘要：The simulation system for navigation satellite signals can simulate the information of multiple satellites under actual conditions， playing a key role in measuring various characteristics of the receiver， reducing the development cost of the receiver， and improving development efficiency. Navigation messages are important basic information for navigation satellite positioning systems， and their accuracy directly affects the accuracy of the positioning system. In this context， the basic structure and content of four GNSS navigation messages were studied， and four GNSS navigation message system simulation modules were constructed successively. The navigation messages of each system were simulated using domestic software MWORKS to obtain binary message information broadcasted during a certain period of time.The decoded data was compared with the original parameters， and the error order of magnitude was much smaller than the original parameters， verified the accuracy of message compilation.关键词：global navigation satellite system;navigation message;coding model;computer simulation;MWORKS
- 71
- |
- 13
- |
- 0
发布时间：2023-10-24 - 摘要：In response to the pre planning problem of production logistics in the rolling mill workshop of China Tobacco Industrial Enterprises， the production equipment and logistics equipment in the rolling mill workshop are modeled and modularized using Modelica language. With a logical relationship class model as the core， a reusable and scalable model library suitable for production logistics in the rolling mill workshop is constructed. At the same time， taking the renovation and upgrading of two production lines in a cigarette factory's cigarette packaging workshop as an example， a production line level system model was built， including a coiling machine， a loading and unloading machine， a packaging machine， and related pallets and automatic guided vehicles. The influence of automatic guided vehicle parameters and tray ratio on production was studied， and optimization parameters were provided to guide the renovation of the production line， avoid the risk of systematic resource waste， and explore a path for subsequent factory level multi domain modeling.关键词：roll-on package workshop;Modelica modeling;production logistics;system planning;simulation application
- 43
- |
- 7
- |
- 0
发布时间：2023-10-24 - 摘要：The application of unmanned aerial vehicles （UAVs） is becoming increasingly widespread， with growing demands for lightweight， cost-effective， and high-performance designs. As structural flexibility increases， multidisciplinary coupling effects become more prominent. Therefore， efficient aerodynamic-structural optimization methods are crucial for enhancing the overall performance of UAVs. This paper combines the aerodynamic vortex lattice method and the structural finite element method to conduct aerodynamic-structural analysis in a loosely coupled manner. Rapid nonlinear sensitivity optimization is achieved using adjoint theory and sequential quadratic programming. For two typical UAV types， long-endurance， and high-maneuverability UAVs， the aerodynamic shape of the wings and parameters of the spatial beam structure are optimized to minimize fuel consumption. The results show that the proposed aerodynamic-structural optimization method can adequately consider the aerodynamic-structural effects of UAVs. Under the condition of satisfactory structural safety， it can enhance overall performance and significantly improve design efficiency.关键词：UAV;aeroelasticity;optimization;adjoint theory;sequential quadratic programming
- 29
- |
- 8
- |
- 0
发布时间：2023-10-24 - 摘要：The integrated energy system realizes the comprehensive control and collaborative optimization of multiple energy flows， which is an inevitable trend in the future development of social energy systems. Studying the dynamic characteristics of integrated energy systems is of great significance for the scientific management of integrated energy systems. Currently， research on modeling and simulation of integrated energy systems has achieved many valuable results. However， there are still many shortcomings in simulating the dynamic transmission process of integrated energy system networks， especially for special flow phenomena， which cannot be well simulated. Therefore， this paper takes the gas-heat-electricity integrated energy system as the research object， adopts a modular modeling method， and establishes an gas-heat-electricity integrated energy system model that fully considers the dynamic characteristics of the network based on Modelica and Julia languages. The fast and slow transient flow characteristics of gas pipelines and phenomena such as "water hammer" and "thermal inertia" during transmission are simulated and verified. Finally， the dynamic response of different energy flows under load changes in the integrated energy system is analyzed by simulation. The results show that the method proposed in this paper has good effects on describing the dynamic characteristics of the network in integrated energy systems and the dynamic response between different energy flows. The research results can support the operation optimization of integrated energy systems and multi-party online energy flow analysis， providing a basis for improving the scientific management capability of integrated energy systems.关键词：integrated energy system;dynamic characteristics of network;Modelica;Julia;modeling and simulation
- 8
- |
- 2
- |
- 0
发布时间：2023-10-24 - 摘要：In response to the problem of cumbersome parameter settings and insufficient visualization of simulation results in the joint simulation process of AMESim and MATLAB/Simulink electromechanical hydraulic systems， a customizable， centralized parameter settings， and visualized simulation results simulation virtual platform technology was studied using LabVIEW based on the standardized interface of FMI （Functional Mock up Interface） and ActiveX technology for human-machine interaction interface design and data exchange. The preliminary application experimental results showed that the virtual platform can easily set parameters and interact with data for joint simulation models， with accurate results and intuitive simulation effects. The simulation report can be automatically output， which is conducive to improving work efficiency.关键词：joint simulation;human-computer interaction;multi-physical domain;virtual platform;data interaction
- 11
- |
- 2
- |
- 0
发布时间：2023-10-24 - 摘要：Reconstruction method of Granger causality network based on GRU network is proposed to address the traditional Granger causality that relies on linear dynamics and cannot meet the needs of nonlinear application scenarios. This method divides the entire network reconstruction into neighbor node selection problems for each target node， constructs a Granger causality model based on GRU network for each target node， introduces a simple gating mechanism to control the update of information in the recurrent neural network， and applies a sparse penalty to the network input weight to extract the Granger causality between nodes. Then integrate each sub network to obtain the final complete causal network structure， and consider using regularization optimization methods during the GRU network modeling and training process. The experiments on linear vector autoregressive， nonlinear vector autoregressive， non-uniformly embedded time-delay vector autoregressive， Lorenz-96 model， and DREAM3 competition dataset show that the proposed network has strong robustness， high effectiveness， and obvious superiority in network reconstruction performance..关键词：network reconstruction;causal inference;recurrent neural network;Granger causality;gated recurrent unit
- 11
- |
- 1
- |
- 0
发布时间：2023-10-24 - 摘要：In order to reduce and eliminate the unreliability of foreign industrial software used in the teaching process of engineering courses at higher engineering institutions， this study takes the integrated undergraduate and postgraduate course "antenna theory and technology" at Harbin Institute of Technology （HIT） as an example， and replaces foreign software with domestic industrial software MWORKS. By comparing the responsiveness of the key user requirements of both foreign and domestic software during the teaching process， the study demonstrates the balanced responsiveness of both software platforms. Furthermore， through the comparison of the predominant form of the software's major operational results， the substitutability of the foreign software is proven. To establish a favorable application ecosystem for domestic software， a case study of ideological education integrated with the "professional + ideological" curriculum is introduced， thereby achieving the combined effect of "professional + ideological + substitution of imported software" in curriculum instruction. This not only replaces the software but also educates potential high-end users of the software. The teaching practice was carried out over two semesters， from autumn 2022 to spring 2023， covering multiple levels of professional students within the antenna course cluster at HIT. The application of domestic software has shown favorable teaching and usage effects， laying the foundation for the application ecosystem of MWORKS as a substitution for foreign software.关键词：industrial software;software substitution;antenna course instruction;curriculum ideology;MWORKS
- 13
- |
- 3
- |
- 0
发布时间：2023-10-24 - 摘要：The task of identifying named entities in folklore texts aims to identify entities from folklore texts and classify them into predefined semantic categories， laying the foundation for the preservation and dissemination of folklore. Folk literature is different from general Chinese corpus in that its text has prominent polysemy and numerous domain nouns， which makes it difficult for conventional named entity recognition methods to accurately and fully identify the entities and their categories present in folk literature texts. To address this issue， a folk literature text named entity recognition model TBERT based on BERT is proposed. This model first integrates the corpus features and entity type features of folk literature texts on the basis of the universal Chinese BERT model； Then， the BiLSTM model is used to further extract sequence dependent features； Finally， combine the label constraint information obtained from the CRF model to output the global optimal result. The experimental results show that this method performs well on the dataset of folk literature texts.关键词：folk literature texts;named entity recognition;Fine-Tune;TBERT-BiLSTM-CRF;feature fusion
- 25
- |
- 2
- |
- 0
发布时间：2023-10-24 - 摘要：Tourist route recommendation refers to planning tourist routes that meet their needs. Existing recommendation methods mostly rely on user ratings， lack analysis of other types of interaction behavior between users and attractions， and lack research on contextual recommendations. To this end， a situational recommendation model for tourism routes based on attention mechanism and portrait is proposed. Firstly， use user profiles and scenic spot profiles to deeply depict tourists and scenic spot resources； Then vectorize the portrait and text， and use two attention mechanisms to assign weights to the interaction behavior feature vectors and text feature vectors， respectively； Finally， use the Deep FM model for scoring prediction. Experiments were conducted on interactive data from the Qunar website dataset， and the results showed that compared with the baseline model， the model improved recommendation accuracy by 3.56%， accuracy by 3.41%， recall by 2.3%， AUC value by 3.44%， and average absolute error by 2.42%， demonstrating its superiority in situational recommendation of tourism routes.关键词：interactive behavior;user portraits;attention mechanisms;contextualized recommendations;Deep FM
- 28
- |
- 2
- |
- 0
发布时间：2023-10-24 - 摘要：The financial panic public opinion analysis method based on the LDA-BiLSTM model is proposed to address the characteristics of strong concealment， fast outbreak speed， and non-standard online language of financial panic public opinion. Using financial industry news websites， forums， Weibo， blogs， etc. as data sources， the LDA method based on part of speech filtering is first used to mine financial hot topics in the data. Then， the BiLSTM model is used to process the short text corpus， and the emotional polarity of netizens towards hot topics is analyzed to identify the content of public opinion warning information. The experiment shows that the accuracy of predicting financial panic public opinion tendencies based on the LDA-BiLSTM model is 92.74%， which can provide information support and public opinion suggestions for managers.关键词：financial panic;latent Dirichlet allocation;bi-directional long short-term memory;public opinion analysis and prediction
- 17
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：In order to improve the accuracy of short-term power load forecasting， this paper proposes a mixed strategy based improved whale optimization algorithm （MSWOA） improved Attention-BiGRU short-term electric power forecasting model. The model first uses bidirectional gating recurrent unit （BiGRU） to extract the information of temporal characteristics of power data in both directions， and introduces Attention mechanism to give different weights to the information of hidden states according to the characteristics of extracted information to increase the influence of important information. To deal with the parameter selection problem of the model， the parameters of the neural network model are automatically selected by the MSWOA algorithm， and the parameters of the network model are optimized to make the optimal prediction effect. And by training and prediction of electric load data， the prediction results are compared with those of BiGRU， Attention-BiGRU， and the whale optimization algorithm （WOA） improved Attention-BiGRU models. The test results show that the prediction accuracy of the optimization model proposed in this paper reaches 98.829%， which has better results compared with the traditional WOA model for the improved Attention-BiGRU network model， and has higher accuracy and stability compared with the neural network model with manually selected parameters.关键词：power prediction;bidirectional gating recurrent unit;attention mechanism;improved whale optimization algorithm
- 21
- |
- 2
- |
- 0
发布时间：2023-10-24 - 摘要：As the basis of natural language processing research， named entity recognition is the main task of recognizing proper nouns such as place names， personal names and organization names in text. Aiming at the shortcomings of BERT model， the paper improved the lightweight model， reduced the training time of the model and improved the feature extraction ability on the premise of ensuring the accuracy. Albert and RoBERTa-WWM were used instead of BERT， and combined with BiLSTM-CRF model， experiments were carried out respectively. The experimental results show that when the cross-layer parameter sharing mechanism is adopted to reduce a large number of parameters， Albert not only has the same evaluation indexes as BERT， but also reduces the use of resources to a great extent. After using dynamic mask and full word masking， F1 value increased by about 5%. RoBERTa's dynamic mask and full word masking are more consistent with the research on Chinese named entity recognition.关键词：named entity recognition;deep learning;sequence labeling;preliminary training model
- 39
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：In a real-world recommendation system， all user data is used to train the recommendation model， but there is a problem of ignoring the temporal order between data and the static learning of user interests. In view of this， a sequence recommendation model combining attention mechanism and GRU recurrent neural network is proposed， which explicitly and dynamically models the user's behavior sequence， mines the user's long-term and short-term preferences， and combines the user's own information to construct an adaptive weighted gating unit to combine long-term and short-term preferences， in order to predict the next behavior that the user may experience. The experimental results on the Amazon dataset show that compared to the current benchmark recommendation models GRU4REC， STAMP， SASREC， etc.， the model has improved at least 14.7% and 8.8% in normalized discounted cumulative gain （NDCG） and hit rate （HIT） indicators， respectively， indicating that the model can more accurately capture user interest.关键词：sequence recommendation;attention mechanism;recurrent neural network;deep learning;long-term and short-term preference
- 36
- |
- 4
- |
- 0
发布时间：2023-10-24 - 摘要：Multimodal speech emotion recognition requires a comprehensive understanding of the content of speech （text information） and the way of speech （acoustic information）， but how to effectively integrate the features of the two modes is a challenging problem. To solve this problem， a multi-modal emotion recognition model based on bidirectional gated recurrent unit （Bi-GRU） and multiple attention is proposed. Firstly， the bidirectional gated recurrent unit is used to extract the features of speech and text modes. Then， the multi-modal feature fusion network composed of parallel self-attention module and guided attention module is used to capture the intermodal and inter-modal interaction relations， so that the model can pay attention to the important intermodal and intra-modal interaction features in the training and learning process. Then the presentation capability of the model is enhanced. The proposed model was evaluated on the IEMOCAP dataset， and the experimental results show that the results of the proposed model are significantly improved compared with other methods.关键词：multimodal emotion recognition;bidirectional circulation gating unit;guide attention module;self-attention module;fusion of features
- 14
- |
- 1
- |
- 0
发布时间：2023-10-24 - 摘要：In order to reasonably evaluate the integrity category of foundation piles， three scale analytic hierarchy process （AHP） is proposed from the three aspects of sound velocity， wave amplitude and waveform respectively through the ultrasonic transmission method of foundation piles. The comparison matrix of evaluation indexes is established to determine the weight of each index feature. The membership matrix is established in combination with the membership function of each evaluation index. The fuzzy comprehensive evaluation model of foundation piles is established. The integrity category of acoustic measuring lines is determined according to the maximum membership principle in fuzzy mathematics. Then， according to the stress structure of the foundation pile， the integrity category of the foundation pile is comprehensively evaluated from the three parts of the acoustic measuring line （detection profile）， the detection cross-section and the whole pile body. Three scale fuzzy analytic hierarchy process is used to judge simulated defective piles in Shandong， and the evaluation result is consistent with the current evaluation standard， reflecting the practicability and feasibility of this method.关键词：pile integrity;ultrasonic transmission method;three scale;fuzzy comprehensive evaluation
- 4
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：To enhance the anti-interference ability and robustness of quadcopter unmanned aerial vehicles under natural wind interference， a control strategy combining backstepping adaptive control with improved fractional order PID control is proposed. Firstly， analyze the mechanical principles of the quadcopter drone and establish a mathematical model of the quadcopter subject to airflow interference； Then the traditional PID controller is improved by incomplete differentiation and feedforward compensation to control the outer loop position subsystem of the quadrotor， and a backstepping adaptive controller is designed to control the inner loop attitude subsystem. At the same time， the stability of the attitude subsystem is verified by Lyapunov stability theory； Finally， simulation experiments were conducted on the Simulink platform. The experimental results show that the controller designed based on the combination of backstepping adaptive control and improved fractional order PID control has more advantages in speed， stability， and accuracy compared to traditional PID controllers and backstepping adaptive controllers.关键词：four rotor UAV;ambient wind interference;incomplete differentiation;feedforward compensation;fractional PID control;backstepping adaptive control
- 6
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：At present， the real-time optimization model of energy management based on the ship power grid model can adjust the output proportion of each distributed power source to improve the comprehensive benefits of the all-electric drive ship during navigation. In order to further improve the efficiency of ships， according to the actual characteristics of the optimization problem， an improved method of honeybadger optimization algorithm （IHBA） is proposed to adjust the density factor calculation method and individual behavior mode selection method. The experiment shows that the comprehensive cost of the energy management strategy obtained by IHBA is always lower than the standard algorithm in the same running time， which proves the effectiveness and progressiveness of the improved method， in order to provide a new solution for the energy management of all electric drive ships.关键词：all-electric ship;energy management strategy;real-time optimization;honey badger algorithm;digital technology
- 6
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：Deeply analyzing the analysis of temperature field distribution in the cooling zone during the firing process of sintered brick kilns is of great significance for the wall material industry to achieve high-quality and high-yield sintered bricks， as well as energy conservation and emission reduction. Firstly， take the stack unit of a certain sintered brick production enterprise as the research object， analyze the heat transfer mechanism of the billet stack unit in the cooling zone， and establish the corresponding mathematical model and geometric model. Then， the model is divided into polyhedral grids， and initial parameters and boundary conditions are set based on enterprise data. Finally， Numerical simulation research on heat transfer in the cooling zone of the tunnel kiln were conducted using Fluent software. The influence of factors such as wind speed and hole shape on the temperature field of sintered bricks was analyzed. The results show that selecting the optimal cooling wind speed can reduce the temperature difference between the inner and outer surfaces of the brick； appropriate hole shape can improve the uniformity of stack temperature distribution； properly increasing the width of the airflow channel between each brick billet can further improve the cooling effect. The research results provide theoretical basis and technical guidance for enterprises to optimize the production process of the cooling section and produce high-quality sintered bricks during the kiln firing process.关键词：tunnel kiln;cooling zone;heat transfer;temperature field;numerical simulation
- 6
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：Intrusion detection systems play an important role in detecting network anomalies and ensuring network security in the power system. Power flow has the characteristics of fixed data flow direction， high recall requirements， and strong timing. To solve the problems of high computational complexity and low accuracy in general detection methods， an efficient power intrusion detection method based on EfficientNet is proposed. This method first preprocesses and converts power flow data； Then， efficient model EfficientNet is used to extract frame level network attack features from input power data；Finally， the extracted image level feature representations are mapped to the classification space， and a fully connected layer network and softmax are used to classify and output detection results， achieving network intrusion detection of power data. The experimental results show that this method can effectively reduce the number of model parameters， reduce model complexity， and improve the efficiency of anomaly traffic intrusion detection while maintaining high classification accuracy.关键词：neural network;electricity flow;intrusion detection
- 14
- |
- 2
- |
- 0
发布时间：2023-10-24 - 摘要：With the development of artificial intelligence， cloud computing， cloud storage， and the Internet of Things technology， smart home， as an emerging industry， has gradually entered people's vision. The smart home system is built on top of the Internet of Things technology， and the system edges are composed of a large number of wireless sensors. Ensuring the integrity of data collected by each edge sensor on the cloud server is an important research topic. The certificate based homomorphic aggregation signature scheme can be easily deployed on smart home systems and can quickly and efficiently achieve batch verification of multi user data integrity in the smart home cloud. The scheme can resist attacks from super attackers in certificate based homomorphic signatures in terms of security， while also preventing collusion attacks from malicious attackers in aggregated signatures. After conducting simulation experiments on the performance of the scheme using the JPBC cryptographic library， the results show that compared to other schemes， this signature scheme has lower computational overhead during signature verification， and can significantly improve the efficiency of cloud based multi user data integrity batch verification in smart home systems.关键词：information security;smart home;sensor networks;aggregate signature;homomorphic signature
- 11
- |
- 3
- |
- 0
发布时间：2023-10-24 - 摘要：Based on the balance sheet data of 151 Chinese banks in 2019， a multi-agent model is used to simulate the risk contagion of changes in interbank bilateral lending relationships and amounts when external loans of banks in different regions are impacted to varying degrees. By proposing a new cross entropy algorithm with local iteration， a lending network with type and regional connectivity preferences is constructed among a large number of banks. The non fully connected lending network between Chinese banks formed based on lending preferences has many clusters. Research has shown that the internal liquidity of banks can conditionally regulate their risk resistance after being impacted， and changes in interbank liquidity can enhance the risk resistance of small banks with low external liquidity or limited lending relationships； Increasing interbank lending can enhance the resilience of large banks with multiple lending relationships and high external liquidity， as well as suppress the scope of bankruptcy risk transmission. The study provides monitoring of internal and external liquidity of banks in various regions， by formulating targeted regulatory policies to seize the timing of regulation and rescue.关键词：interbank lending network;connection tendency;internal and external liquidity;multi-agent;risk contagion
- 4
- |
- 1
- |
- 0
发布时间：2023-10-24 -
Research on the Promotion Mechanism of Incentive for the Adoption of Face Recognition Technology AI导读
摘要：To explore whether incentive measures have a promoting effect on the adoption rate of facial recognition technology， a technology acceptance model was adopted， and incentive mechanism theory was introduced. A model was constructed combining perceived usefulness， perceived ease of use， and adoption behavior. AMOS was used to conduct hypothesis testing and mediating effect analysis on the constructed model. The results indicate that economic incentives have a promoting effect on the perceived ease of use of facial recognition； Power incentives have a promoting effect on the perceived ease of use and perceived usefulness of facial recognition； Social motivation has a promoting effect on the perceived usefulness and perceived ease of use of facial recognition； The perceived usability of facial recognition has a positive impact on perceived usefulness and adoption behavior； The perceived usefulness of facial recognition has a positive impact on adoption behavior. The research results demonstrate the mechanism of economic incentives， power incentives， and social incentives on the adoption of facial recognition technology， providing practical guidance for adopting different incentive measures to promote the adoption of facial recognition technology.关键词：face recognition;technology adoption;economic incentive;power incentive;social incentive- 4
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：Based on the tendency of depression patients to express emotions on social media， it is proposed to search， analyze， and discuss the main characteristics of depression patients on current social media. Firstly， high-frequency word analysis was conducted on Weibo data of depression patients， and mapping relationships were constructed using LDA topic models； Then， based on time series analysis， the changes in positive and negative emotional expression of this group of people are analyzed， and the intensity proportion of five types of negative emotions is analyzed in detail； Finally， based on existing theoretical achievements， summarize the platform image， cognitive characteristics， behavioral characteristics， and emotional characteristics of depression patients. Identifying potential depression patients using text features and key emotional influencing factors in Weibo platform have significant practical significance.关键词：depression;social media;text semantic mining;emotional analysis
- 10
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：Aiming at the cold start problem of collaborative filtering algorithm and lack of book evaluation data in university libraries， a popular book recommendation method based on analytic hierarchy process is proposed. Firstly， a book popularity evaluation model is constructed based on the analytic hierarchy process method， then books’ similarities are calculated by extracting Chinese library classification （CLC） numbers， finally Top-N popular book lists and Top-N new book lists are recommended based on similar CLC numbers.Through experimental verification and comparison on the dataset of university libraries， the results show that the recommendation diversity and novelty of the proposed algorithm are better than the traditional algorithms， and the recommendation performance is improved.关键词：university libraries;collaborative filtering algorithm;analytic hierarchy process;Chinese library classification number;book popularity evaluation
- 12
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：The high similarity between fine-grained clothing image categories makes their retrieval challenging in imbalanced datasets. Therefore， a cost sensitive fine-grained clothing image retrieval method is proposed. By designing a cost sensitive loss function to deal with unbalanced data sets， the key point detection module of the retrieval model is effectively improved to improve the fine-grained clothing image retrieval performance. The cost sensitive loss function strategy is based on the category rebalancing of fixed weighting and dynamic weighting， where the fixed weighting is based on the prediction probability assignment of category frequency and labels， while the dynamic weighting adjusts its weight according to the prediction score， allowing the model to adjust instances of different difficulties， thus improving the gradient update weight of difficult class samples. The ablation experiment of the cost sensitive loss function on the clothing dataset DeepFashion shows that both fixed weighting and dynamic weighting improve the model retrieval performance under unbalanced data sets. Compared with other fine-grained image retrieval methods， the experiment further shows that the cost sensitive loss function can solve the problem of category imbalance and difficult category retrieval. In addition， the experimental results of clothing attribute retrieval such as category and style show that the improvement of the proposed model and the optimization of loss function are effective.关键词：fine-grained clothing;image retrieval;cost-sensitive;imbalanced dataset
- 7
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：To address the issues of false detection， missed detection， and low accuracy of current highway cracks， an improved road crack recognition model based on YOLOv5 is proposed. Firstly， the feature extraction ability of the algorithm is improved by adding attention module （CBAM） to the backbone network； Then， modify the network feature fusion method and replace the feature pyramid （FPN）+pixel aggregation network （PAN） structure with a weighted bidirectional feature pyramid （BiFPN） to enhance feature fusion. Finally， an attention mechanism is added at the tail of the weighted bidirectional feature pyramid and the head of the detection layer to enhance the network's expression ability. The experimental results indicate that the improved model outperforms the traditional YOLOv5 model in terms of mAP@0.5 and mAP@0.5∶0.95 by 33.5% and 21.5%， respectively. The detection effect of improved YOLOv5 algorithm for road cracks is good， and it can quickly and accurately identify and locate cracks on high-speed roads.关键词：road disease;YOLOv5;CBAM attention mechanism;feature fusion;BiFPN
- 56
- |
- 1
- |
- 0
发布时间：2023-10-24 - 摘要：Rainy weather will cause rain marks on the image， which will not only seriously affect the look and feel of the image， but also interfere with subsequent image analysis and processing. Image rain removal has always been the focus of image restoration research. In view of this， a single image rain removal model of RDSRCNN is proposed. In order to improve the feature extraction capability， ESIFEM， an enhanced feature extraction method， was used as the feature extraction means to achieve efficient feature extraction by utilizing its remote pixel correlation capability and low local feature extraction cost. Meanwhile， the loss function constructed with the combination of the l1 and MSS-SIM was used to optimize the training efficiency and ensure the visual friendliness of the output image. The above method is combined with the enhanced DSRCNN rain removal network to form a single image rain removal model. The experimental results on the Rain100H dataset show that this method can visually restore the image with dense rain distribution to a clean scene map with rich details， and compared with most comparison methods， the cases of ghost and object edge deformation are reduced by more than 90%， and the clearance rate of background rain marks is higher than 95%. In quantitative evaluation， the proposed method is superior to the comparison method in most aspects of peak SNR and structural similarity parameters and is superior to Restormer method in space complexity.关键词：image inpainting;single image de-rain;deep learning;mixed loss function
- 9
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：In the visual SLAM （simultaneous localization and mapping）， the method of using semantic segmentation and object detection to detect dynamic objects and remove outliers has become the mainstream， but its disadvantage is that it is unable to fully track the semantic information of objects. Therefore， this paper proposes an improved semantic SLAM algorithm based on object tracking， which uses YOLACT++network to segment object mask， extract object feature points， and use inter frame matching to achieve object tracking. The method detects the depth， reprojection error and epipolar constraint of the matched feature points， and then judges the dynamic and static state of the object to achieve object tracking and judge the motion state. After testing the TUM RGB-D dataset， the experiment shows that the method can effectively track objects， and the trajectory estimation accuracy is better than other SLAM algorithms， which has practical value.关键词：visual SLAM;semantic segmentation;object tracking;dynamic environment;geometric constraint
- 9
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：Unified modeling language （UML） is an important course in the software engineering curriculum system. Driven by engineering education certification and talent demand in the new era， this course should improve the adaptability of engineering education talent training for the industrial development. Although the existing UML curriculum reform has been improved via the case analysis， mixed teaching， and other methods， there are still shortcomings in combining "learning by doing" and "team cooperation case teaching". Starting from the sensitivity of students majoring in software engineering to programming， and guided by the SE-CDIO engineering education concept， this paper organically integrates IBM-RSA and PlantUML command line modeling， case inspiration， and online teaching assistance into the SE-CDIO method， and proposes a UML course teaching method based on "SE-CDIO-CA". This method starts from the training objectives and graduation requirements of software engineering， and combines with the whole cycle of software life， moreover， the proposed method is specifically designed from four aspects： project requirements analysis and modeling， project design and optimization， project implementation and optimization， and project maintenance and release. The whole teaching process closely focuses on the basic concept of "engineering practice teaching"， and constantly improves students' interest and ability to solve practical engineering problems using UML theoretical knowledge.关键词：concept of engineering education;UML course;cycle of software life;teaching of engineering practice;software engineering
- 11
- |
- 1
- |
- 0
发布时间：2023-10-24 - 摘要：A semi physical simulation teaching and training platform has been developed based on the actual teaching of land air communication courses， combining voice， image， and animation to achieve dynamic， image based， digital， and audio based teaching. This platform can complete the learning and training of various units of the land air communication course， achieve multiple teaching modes， and improve students' participation enthusiasm and confidence in learning the course well. Platform based teaching has expanded the breadth and depth of language teaching training， improved students' thinking level and comprehensive ability， improved teaching efficiency and quality， and injected vitality into the teaching of land and air communication courses through its process oriented training methods and diversified assessment methods.关键词：radio communication for pilots;Unity 3D;virtual simulation;process assessment
- 3
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：In view of the shortcomings in the teaching process of centralized practical courses for software engineering majors， the underlying causes and pain points that need to be solved are analyzed in this paper， the educational elements of innovation， entrepreneurship and creation are introduced， with the needs of engineering education certification， operational teaching reform measures are proposed. Taking "Software quality assurance and testing course design" as an example， the exploration process of teaching reform is expounded from the aspects of re-defining the observation points of graduation requirements， re-improving the way of teamwork， re-selecting targeted assessment contents， re-formulating multiple assessment and evaluation methods， etc.， and finally the achieving degree of the course is calculated， analyzed and evaluated to clarify the effectiveness of the teaching reform.关键词：software engineering;centralized practical course;“three innovations” education;engineering certification;achieving degree analysis
- 8
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：Digital resources are a hot topic in the field of education research， and their content and form are still in the ascendant after more than 20 years of development. New concepts and technologies that are emerging in response to the times continuously improve relevant theoretical and practical research. Differentiate and analyze the concept and connotation of digital resources， sort out domestic literature， use CiteSpace software to conduct metrology analysis on 1 641 extracted literature， and display its keyword clustering map and prominence word analysis map； At the same time， the research content in the field of digital education resources in China was analyzed and organized， the existing problems were analyzed， and corresponding development suggestions and countermeasures were proposed.关键词：digital education resources;CiteSpace;bibliometric analysis;cluster analysis
- 10
- |
- 0
- |
- 0
发布时间：2023-10-24 - 摘要：As a visual representation of knowledge， educational knowledge graph forms a structured knowledge system by processing massive disordered and complicated data， and constantly excavates new development space in the field of education. Through systematic literature review， systematically comb the empirical research results of the past 10 years in China to reveal the specific types， applications and challenges of knowledge map in the field of education. The research found that the educational knowledge map mainly includes five types： subject knowledge map， group knowledge map， multimodal knowledge map， learning cognitive map and educational knowledge base； Knowledge atlas is widely used in five aspects： learner portrait construction， learning situation diagnosis and learning evaluation， learning resource recommendation， personalized learning path planning， intelligent management and knowledge question and answer system. In addition， in the construction and application process of education knowledge map at the current stage， there are still practical challenges such as low quality of education resource data sets， difficult integration of multi-source knowledge map， and single evaluation method of education knowledge map.关键词：knowledge graph;educational knowledge graph;systematic literature review;construction of knowledge
- 15
- |
- 3
- |
- 0
发布时间：2023-10-24 - 摘要：Android system is currently the open-source mobile operating system with the highest market share， but its open-source nature makes it the main target of malicious software attacks. Malware poses a serious threat to national security and personal privacy， and its evasion behavior is becoming increasingly covert， making it often difficult to detect and analyze. To this end， first summarize the types of malicious application features extracted using static and dynamic analysis methods for Android malicious application detection； Then， from the perspectives of traditional machine learning and deep learning， summarize and analyze Android malicious application detection methods， and summarize the work related to Android malicious application detection and defense from the perspectives of adversarial attacks and protection； Next， compare the intelligent analysis methods for Android malicious applications on a common dataset； Finally， the research directions and challenges of intelligent analysis methods for Android malicious applications in the future are discussed and summarized from aspects such as features， datasets， and analysis models， in order to provide reference for their development.关键词：Android malicious application;malware detection;machine learning;deep learning;adversarial attack
- 12
- |
- 1
- |
- 0
发布时间：2023-10-24
- 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