Top Read Articles

Published in last 1 year |  In last 2 years |  In last 3 years |  All

Please wait a minute...


For Selected: Download Citations EndNote Ris BibTeX Toggle Thumbnails

Select

Review of Research on Human Behavior Detection Methods Based on Deep Learning SHEN Jia-wei, LU Yi-ming, CHEN Xiao-yi, QIAN Mei-ling, LU Wei-zhong, Computer and Modernization    2023, 0 (09): 1-9.   DOI: 10.3969/j.issn.1006-2475.2023.09.001 Abstract （501）      PDF（pc） （2112KB）（139）       Knowledge map    Save Human behavior recognition has always been a hot topic of research in the field of computer vision and video understanding and is widely used in other areas such as intelligent video surveillance and human-computer interaction in smart homes. While traditional human behavior detection algorithms have the disadvantages of relying on too many data samples and being susceptible to environmental noise， evolving deep learning techniques are gradually showing their advantages and can be a good solution to these problems. Based on this， this paper firstly introduces some commonly used behavioral recognition datasets and analyses the current research status of human behavioral recognition based on deep learning， then describes the basic process of behavioral recognition and commonly used behavioral recognition methods， finally summarizes the performance， existing problems of various existing behavioral recognition methods， and outlooks the future development directions. Reference | Related Articles | Metrics | Comments（0）

Select

Feature-level Multimodal Fusion for Depression Recognition GU Ming-xuan, FAN Bing-bing Computer and Modernization    2023, 0 (10): 17-22.   DOI: 10.3969/j.issn.1006-2475.2023.10.003 Abstract （350）      PDF（pc） （1213KB）（175）       Knowledge map    Save Abstract: Depression is a common psychiatric disorder. However， the existing diagnostic methods for depression mainly rely on scales and interviews with psychiatrists， which are highly subjective. In recent years， researchers have devoted themselves to identifying depressed patients by EEG features or audio features， but no study has effectively combined EEG information with audio information， ignoring the correlation between audio and EEG data. Therefore， this study proposes a feature-level multimodal fusion model to improve the accuracy of depression recognition. We combine the audio and EEG modality information based on a fully connected neural network. Our experiments show that the accuracy of depression recognition using feature-level multimodal fusion model on the MODMA dataset reaches 81.58%， which is higher than that of using single-modality. The results indicate that the feature-level multimodal fusion model can improve the accuracy of depression recognition compared to single-modality. Our research provides a new perspective and method for depression recognition. Reference | Related Articles | Metrics | Comments（0）

Select

Review of Infrared Small Target Detection HU Rui-jie, CHE Dou Computer and Modernization    2023, 0 (08): 79-86.   DOI: 10.3969/j.issn.1006-2475.2023.08.013 Abstract （308）      PDF（pc） （5630KB）（152）       Knowledge map    Save bstract： This article aims to review three infrared small target detection methods based on traditional feature extraction， local comparison， and widely used deep learning today. Then， by comparing the cutting-edge applications of these three methods， their advantages and disadvantages in target detection performance， robustness， and real-time performance are analyzed. We find that feature extraction based methods exhibit good real-time and robustness in simple scenarios， but may have limitations under complex conditions. The method based on local comparison is relatively robust to changes in object size and shape， but sensitive to background interference. The method based on deep learning performs well in object detection performance， but requires large-scale data and larger computing resources. Therefore， in practical applications， the advantages and disadvantages of these methods should be comprehensively considered based on specific scenario requirements， and appropriate methods should be applied to infrared small target detection. Related Articles | Metrics | Comments（0）

Select

Aspect Based Sentiment Analysis Model Based on Knowledge Enhancement LI Shi-yue, MENG Jia-na, YU Yu-hai, LI Xue-ying, XU Ying-ao Computer and Modernization    2023, 0 (10): 1-8.   DOI: 10.3969/j.issn.1006-2475.2023.10.001 Abstract （245）      PDF（pc） （2224KB）（100）       Knowledge map    Save Aspect based sentiment analysis can accurately determine the emotional polarity of aspect words in sentences， and plays an important role in social networking， e-commerce and other fields. Most of the existing methods model the relationship between context and target words through sequence representation or attention mechanism， but ignore the background knowledge of text and the conceptual links between aspect words， resulting in insufficient semantic relationships learned. To solve the above problems， the Aspect Based Sentiment Analysis Model Based on Knowledge Enhancement （ABSA-KE） is proposed. First， the features are extracted and the corresponding word vector is obtained through the pre-training model BERT， and the dependency tree corresponding to the text is obtained using the parser. Then， the joint modeling of BiLSTM and graph attention network is used to learn the node embedded representation and obtain the text vector. Second， the external knowledge base is used to introduce the aspect word knowledge vector in different contexts to enhance the aspect level emotion analysis model， and finally the emotion classification task is carried out. Compared with the existing models， the experimental results show that the proposed model is effective and reasonable in aspect level emotion analysis tasks. Reference | Related Articles | Metrics | Comments（0）

Select

Joint Extraction Method of Entities and Relations Based on FGM and Pointer Annotation LIU Yu-peng, GE Yan, DU Jun-wei, CHEN Zhuo Computer and Modernization    2023, 0 (11): 1-5.   DOI: 10.3969/j.issn.1006-2475.2023.11.001 Abstract （212）      PDF（pc） （1192KB）（96）       Knowledge map    Save Abstract: Joint extraction of entities and relations is an important task of information extraction. The traditional entity relationship joint extraction method cannot solve the problem of overlapping triples well， because it models the relationship between entities as discrete types. In order to solve the problem that it is difficult to extract overlapping triples， this paper proposes a BERT-FGM model for entity relationship joint extraction， which combines FGM and pointer annotation. In this model， the relationship between entities is modeled as a function， and the robustness of the model is improved by incorporating FGM into the process of BERT training word vector. The model firstly extracts the subjects through the pointer annotation strategy， then fuses the subjects into a sentence vector as a new vector， and finally uses it to extract objects under a predefined relationship condition. Experiments are carried out on public dataset WebNLG， the experimental result shows that the F1 value of the model is 90.7%， it can effectively solve the problem of relationship triples overlapping. Reference | Related Articles | Metrics | Comments（0）

Select

Research on Stock Classification and Forecast Based on DTW-TCN SUN Zi-yu, REN Ran, WEI Xi-zhe Computer and Modernization    2023, 0 (08): 31-37.   DOI: 10.3969/j.issn.1006-2475.2023.08.006 Abstract （186）      PDF（pc） （8512KB）（96）       Knowledge map    Save Abstract: With the development of society and information technology， financial instruments and stock transactions have taken on a new form， namely， the number of financial data increases. Therefore， stock trend prediction is particularly important in high-frequency trading. Stock trend prediction in high-frequency trading is particularly important to improve the accuracy of stock trend prediction in high-frequency trading. A temporal convolutional network （TCN） model based on dynamic time warping （DTW） clustering analysis is proposed. In the model， the opening price， the highest price， the lowest price， the closing price， the trading volume， and the trading volume are used as the stock characteristic variables. In order to avoid the influence of magnitude， the feature vector is standardized first， and then the stock is classified by using the dynamic time warping to measure the similarity of time series， Then， temporal convolutional network （TCN） extracts the common characteristics of the categories to predict the opening and closing price trends of the stocks of the categories， and compares them with the actual trends. The experiment is conducted with the minute-level data of 19 industry universal stocks. Compared with traditional time series model and LSTM network model， it has greater time characteristics. The results show that the model can effectively classify the stocks with the same trend into the same category， and achieve accurate trend prediction in the minute-level high-frequency trading. Related Articles | Metrics | Comments（0）

Select

A Fast Registration Method for Massive Point Clouds Based on 3D-SIFT and 4PCS LI Jia-le1, LI Zhe-run1, ZHAO Yong2, ZHANG Yang1 Computer and Modernization    2024, 0 (02): 1-6.   DOI: 10.3969/j.issn.1006-2475.2024.02.001 Abstract （175）      PDF（pc） （1952KB）（294）       Knowledge map    Save Abstract： The registration of measurement point cloud and model point cloud is the key of visual positioning. Aiming at the problems of poor visual positioning accuracy and low algorithm efficiency caused by large amount of measurement point cloud data and low overlap rate with CAD model point cloud， a registration method of measurement point cloud and model point cloud based on the fusion of 3D scale invariant feature transform （3D-SIFT） and four point fast robust matching algorithm （4PCS） is proposed. Firstly， the depth camera is used to extract the point cloud of the part， and the extracted measurement point cloud is denoised and filtered； Then 3D-SIFT feature point extraction algorithm is used to extract feature points from measurement point cloud and CAD model point cloud； Finally， the extracted feature points are used as the initial values of the 4PCS algorithm to achieve the registration of the two point cloud data. Compared with the commonly used 4PCS algorithm and Super-4PCS algorithm， the algorithm simulation and experimental results show that the proposed algorithm can improve the registration speed by more than 30% on the premise of ensuring the registration accuracy. Reference | Related Articles | Metrics | Comments（0）

Select

Safety Helmet Detection Based on Lightweight YOLOv5 LI Yan-man, WANG Bi-heng, ZHAO Ling-yan Computer and Modernization    2023, 0 (10): 59-64.   DOI: 10.3969/j.issn.1006-2475.2023.10.009 Abstract （173）      PDF（pc） （5102KB）（86）       Knowledge map    Save There is a large amount of data in the intelligent monitoring system of distribution network， which objectively requires the algorithm to have high real-time performance. Based on this， the YOLOv5 algorithm is improved in light weight， including improving the K-means algorithm clustering anchor box， using the Hard-swish activation function and the CRD loss function， and at the same time integrating the ShuffleNet structure in the backbone network and adopting the Attention mechanism in the FPN module. The presented model， SNAM-YOLOv5 （ShuffleNet and Attention Mechanism-You Only Look Once version 5）， can significantly improve the detection performance and the processing speed of small targets and occluded targets. The results of safety helmet detection based on HiSilicon Hi3559A embedded platform show that the model is superior to similar algorithms and has good real-time performance. Reference | Related Articles | Metrics | Comments（0）

Select

Hippocampus Segmentation Based on Feature Fusion CHEN Jia-min, ZHANG Bo-quan, MAI Hai-peng Computer and Modernization    2023, 0 (08): 1-6.   DOI: 10.3969/j.issn.1006-2475.2023.08.001 Abstract （169）      PDF（pc） （1332KB）（68）       Knowledge map    Save Abstract: Aiming at the problem that the existing hippocampal segmentation algorithm can not segment the target accurately， a novel hippocampal segmentation model based on feature fusion using codec structure is studied. Firstly， Resnet34 is used as the model feature encoding layer to extract richer semantic features; Secondly， the ASPP module based on mixed expansion convolution is introduced into the coding and decoding transition layer to obtain multi-scale feature information. Finally， the attention feature fusion mechanism is used as the connection layer between the encoding and decoding layers to effectively combine the deep features with the shallow features， provide the location information of the hippocampus for subsequent segmentation， and improve the segmentation performance of the model. The experiment is carried out on ADNI dataset to verify the validity of the proposed model. The accuracy of the network model in the four evaluation indicators of IoU， DICE， accuracy and recall rate reaches 84.67%， 88.51%， 87.90% and 89.01% respectively. Compared with the existing advanced medical segmentation algorithm， the experimental results also show that the model has better segmentation ability and achieves better automatic segmentation effect of hippocampus image. Related Articles | Metrics | Comments（0）

Select

Prediction of Diabetes Mellitus Using LightGBM Classifier with RF-RFECV LIU Jing-le, LUO Xiang, GONG Cheng-rong, ZHANG Guo-peng Computer and Modernization    2023, 0 (11): 36-43.   DOI: 10.3969/j.issn.1006-2475.2023.11.006 Abstract （162）      PDF（pc） （2220KB）（95）       Knowledge map    Save Abstract: In order to find the high-risk population of diabetes in China as early as possible and provide targeted intervention measures， the data set of China Health and Retirement Longitudinal Study （CHARLS）， which represents the Chinese population， was selected as the research object， and a hybrid algorithm based on RF-RFECV and LightGBM （RF-RFECV-LightGBM） was proposed， and compared with five other algorithms through experiments. The results show that RF-RFECV- LightGBM has the best overall performance， the accuracy， precision， recall， F1 value and AUC value are 0.9772， 0.9952， 0.8178， 0.8978， and 0.9357， respectively. The prediction time is 0.0428 s， which is 0.0549 s shorter than the prediction time of LightGBM before feature selection （increased by 56.19%）， indicating the effectiveness of RF-RFECV algorithm. Finally， the same prediction process was tested on the Pima Indian dataset， and the results achieved an accuracy of 0.9415， further verifying the excellent performance of the proposed algorithm RF-RFECV-LightGBM， which can assist in clinical diagnosis and treatment of diabetes. Reference | Related Articles | Metrics | Comments（0）

Select

Improved YOLOv7 Algorithm for Low-resolution Ship Object Detection in Complex Backgrounds#br# YAN Zi-xian, DONG Bao-liang, TANG Si-mi Computer and Modernization    2023, 0 (11): 120-126.   DOI: 10.3969/j.issn.1006-2475.2023.11.019 Abstract （159）      PDF（pc） （3631KB）（109）       Knowledge map    Save Abstract: In response to the problems of low resolution target detection and interference from complex backgrounds in ship image target detection， an improved YOLOv7 algorithm is proposed for identifying ship targets. The algorithm is mainly improved in three aspects: using K-means++ algorithm for anchor box clustering in the ship target dataset to obtain anchor box information that is more suitable for ship detection tasks; improving the loss function by using EIOU loss instead of CIOU loss and using Focal loss combined with ɑ-Balanced instead of standard cross-entropy loss; improving the network structure by adding the SPD-Conv module to enhance the detection effect for low-resolution targets. Experimental results show that compared with the original YOLOv7 algorithm， the improved algorithm has an accuracy improvement of 4.22 percentage points， a recall rate improvement of 2.68 percentage points， a mAP@0.5 improvement of 4.3 percentage points， and a detection speed improvement of 2 frames/s. The algorithm achieves good detection results for ship targets. Reference | Related Articles | Metrics | Comments（0）

Select

Cross Modal Hash Retrieval Based on Attention Mechanism and Semantic Similarity WANG Hong, GE Hong Computer and Modernization    2023, 0 (08): 44-53.   DOI: 10.3969/j.issn.1006-2475.2023.08.008 Abstract （150）      PDF（pc） （2748KB）（82）       Knowledge map    Save Abstract： Nowadays， cross-modal hash retrieval has been widely and successfully used in multimedia similarity search applications. There are two challenged questions in deep hash retrieval methods：1）How to measure multiple modal’s similarity more accurately. 2）How to fuse multiple modal’s features to gain more abundant feature representations， so as to avoid key information loss. Therefore， in order to solve these two problems， we propose a novel cross-modal hashing method， called cross-modal hash retrieval model based on attention mechanism and semantic similarity （ASSH）， by defining a new multi-label similarity measurement method to distinguish the importance of different labels， designing an attention fusion module to fuse the features and enhance the interaction between different modal. Experimental results demonstrate that the proposed method outperforms the previous methods in all problem modes on the three common datasets MIRFLICKR-25K， NUS-WIDE and IAPR TC-12. Compared to the state-of-the-art method， when the hash code length is 16 bit， the mean Average Precision （mAP） is improved by 1.1% and 0.63%. At the same time， the ablation experiment also fully proved the effectiveness of the method. Related Articles | Metrics | Comments（0）

Select

Real-time Detection of Arbitrary Shape Scene Text Based on Segmentation XU Hong-kui, LI Zhen-ye, GUO Wen-tao, ZHAO Jing-zheng, GUO Xu-bin Computer and Modernization    2023, 0 (11): 95-100.   DOI: 10.3969/j.issn.1006-2475.2023.11.015 Abstract （147）      PDF（pc） （1710KB）（85）       Knowledge map    Save Abstract：The current challenges of scene text detection technology are mainly reflected in two aspects: the trade-off between model real-time performance and accuracy， and the detection of arbitrary shape text. They determine whether scene text detection is feasible in real scenes. Aiming at the above two problems， this paper proposes a lightweight backbone network with strong feature extraction ability based on segmentation method， which can accurately detect natural scene text of arbitrary shape in real time. Specifically， a simple dual-resolution residual backbone network and a deep aggregate pyramid pooling module with low computational cost are used， and the features extracted from them are fused and segmented using a differentiable binarization module. Through the comparative experiment on the standard English dataset ICDAR2015， the result show that the improved method proposed in this paper is effective， and achieves comparable results in real-time performance and accuracy. Reference | Related Articles | Metrics | Comments（0）

Select

Survey of Supply Chain Oriented Consensus Algorithms CHAI Li, WANG Xiao, GONG Jia-hao, WANG Yang, JI Shun-hui, ZHANG Peng-cheng Computer and Modernization    2023, 0 (11): 22-27.   DOI: 10.3969/j.issn.1006-2475.2023.11.004 Abstract （140）      PDF（pc） （993KB）（84）       Knowledge map    Save Abstract： As one of the core technologies in the blockchain， consensus algorithm is an important method for the system to maintain data consistency and security. This paper firstly investigates and analyzes the relevant research on the universal consensus algorithms in the alliance chain， classifies algorithms from the perspective of whether they are based on the Byzantine problem， and combs and summarizes consensus algorithms from four aspects： problem entry， principle description， performance analysis and application scenarios. In addition， focusing on the application scenarios related to the supply chain， this paper analyzes the challenges it brings to the consensus algorithm in the alliance chain， and sorts out and summarizes the consensus algorithms in the alliance chain under this scenario. Finally， the paper discusses the challenges faced by the consensus algorithm and the direction for future development， with an intention of providing references for researchers in this field. Reference | Related Articles | Metrics | Comments（0）

Select

Sequence Recommendation Model Based on Dynamic Convolution and Self-attention ZHENG Hai-li, CHEN Ping-hua Computer and Modernization    2023, 0 (10): 9-16.   DOI: 10.3969/j.issn.1006-2475.2023.10.002 Abstract （139）      PDF（pc） （3775KB）（65）       Knowledge map    Save Sequence recommendation dynamically models user interests according to the historical interaction records of users and items， and recommends next item. The sequence modeling user interests is usually divided into long-term interest dependency and short-term interest dependency. The existing methods either divide the sequence according to the interaction order， respectively model the long-term and short-term interest dependence， separately model the user interest， or extract the features of the same interactive sequence in parallel with different feature extraction technologies to obtain the global and local interest representation， ignoring the fact that the user intention at different times exists in the behavior context at that time. This paper proposes DConvSA to model dynamic interest by using dynamic convolution and self-attention. Dynamic convolution is used to extract local dynamic interest， and convolution kernel is generated according to different context items to adaptively calculate the importance of items. Combined with the self-attention mechanism， the overall significant item dependency is obtained. At the same time， the global and local interest dependencies at each time are fused in an explicit way to better model the relationship between interests at different times. Experiments are conducted on three public datasets， using recall rate， average reciprocal ranking and normalized cumulative gain for performance evaluation. The results show that the recall rate， mean reciprocal ranking and normalized discounted cumulative gain increased by at least of 1.53%， 3.77% and 3.28% on the MovieLens-1M dataset， 1.86%， 1.94% and 2.46% on the Amazon Beauty dataset， and 0.22%， 0.97% and 1.08% on the Steam dataset. Reference | Related Articles | Metrics | Comments（0）

Select

Image Super-resolution Reconstruction Based on Spatial Attention Residual Network XING Shi-shuai, LIU Dan-feng, WANG Li-guo, PAN Yue-tao, MENG Ling-hong, YUE Xiao-han Computer and Modernization    2023, 0 (10): 45-52.   DOI: 10.3969/j.issn.1006-2475.2023.10.007 Abstract （136）      PDF（pc） （3939KB）（88）       Knowledge map    Save  Hierarchical features extracted from convolutional neural networks contain affluent semantic information and they are crucial for image reconstruction. However， some existing image super-resolution reconstruction methods are incapable of excavating detailed enough hierarchical features in convolutional network. Therefore， we propose a model termed spatial attention residual network （SARN） to relieve this issue. Specifically， we design a spatial attention residual block （SARB）， the enhanced spatial attention （ESA） is embedded into SARB to obtain more effective high-frequency information. Secondly， feature fusion mechanism is introduced to fuse the features derived from each layer. Thereby， the network can extract more detailed hierarchical features. Finally， these fused features are fed into the reconstruction network to produce the final reconstruction image. Experimental results demonstrate that our proposed model outperforms the other algorithms in terms of quantitative evaluation and visual comparisons. That indicates our model can effectively utilize the hierarchical features contained in the image， thus achieve a better super-resolution reconstruction performance. Reference | Related Articles | Metrics | Comments（0）

Select

Apple Defect Detection Algorithm Based on NAM-YOLO Network ZHANG Jia-Qi, XU Qi-lei Computer and Modernization    2023, 0 (10): 53-58.   DOI: 10.3969/j.issn.1006-2475.2023.10.008 Abstract （135）      PDF（pc） （4281KB）（96）       Knowledge map    Save Aiming at the problems of apple defect detection， such as frequent false detection， leakage detection and easy confusion of defects， we propose an apple defect detection algorithm based on improved YOLOv5. Apple defect detection is very important for apple sorting. The existing methods of apple defect detection mainly extract color and texture features through machine learning or convolutional neural network， but there are problems such as error detection， missing detection and insufficient feature extraction ability. It can not meet the requirements of accuracy and real-time defect detection. NAM-YOLO algorithm mainly has three core ideas: 1） By adding TRANS module to the backbone network， features and global information can be better integrated； 2） The weighted bidirectional feature pyramid network is used to fuse features of different scales； 3） The NAM attention mechanism based on normalization is introduced into the neck network to strengthen the key features of the target region and improve the detection accuracy of the network. Experimental results show that the mAP of the improved algorithm reaches 98.90% and the accuracy is 98.73%. Compared with other models， this model has better feature fusion ability and can better meet the actual needs of apple sorting. Reference | Related Articles | Metrics | Comments（0）

Select

Incremental News Recommendation Method Based on Self-supervised Learning and Data Replay LIN Wei Computer and Modernization    2023, 0 (12): 1-6.   DOI: 10.3969/j.issn.1006-2475.2023.12.001 Abstract （127）      PDF（pc） （1490KB）（105）       Knowledge map    Save Abstract： Personalized news recommendation technology is important to alleviate information overload and improve user experience. News recommendation models are usually iteratively trained based on fixed data sets. However， in real scenarios， news recommendation models need to constantly learn to adapt to new users and news. Therefore， incremental learning has been proposed to help models perform incremental updates. The main challenge of the incremental learning of news recommendation models is the catastrophic forgetting problem， where a trained model forgets the user preferences it has previously learned. In view of this， this paper proposes SSL-DR， an incremental learning method of news recommendation models based on self-supervised learning and data replay. SSL-DR firstly adds the self-supervised learning task to the news recommendation task to obtain the user's stable preference， which effectively reduces the problem of catastrophic forgetting. To consolidate the learned knowledge， SSL-DR further implements a sampling strategy based on the user's click probability scores for candidate news to achieve data replay and transfer the learned knowledge through a knowledge distillation strategy. The experimental results show that， our method can effectively improve the overall recommendation performance of the news recommendation model in the process of incremental training， and significantly alleviate the problem of catastrophic forgetting. Reference | Related Articles | Metrics | Comments（0）

Select

Breast Cancer Prediction and Feature Analysis Model Based on CatBoost and SHAP JIA Xiao-yao, Computer and Modernization    2023, 0 (10): 32-38.   DOI: 10.3969/j.issn.1006-2475.2023.10.005 Abstract （124）      PDF（pc） （3562KB）（81）       Knowledge map    Save To address the problems of insufficient performance and poor interpretability of current breast cancer prediction models， this paper proposes a breast cancer prediction and feature analysis model incorporating CatBoost and SHAP. First， the original breast cancer dataset is processed with outliers and data normalization to improve the quality of the data. Then， a model for breast cancer prediction based on CatBoost is built and generalization ability analysis is performed. Finally， the prediction model is combined with SHAP for interpretable analysis to explore the key factors affecting breast cancer. The model is validated using the Breast Cancer Wisconsin （Diagnostic） dataset from the University of Wisconsin， and the results show that the Accuracy value of 99.30%， Precision value of 99.50%， Recall value of 98.91%， and F1 value of 99.19% are better than the existing literature. The superiority of this model is verified by the fact that the Accuracy index improved by 1.12-6.90 percentage points， the Precision index improved by 2.00-7.50 percentage points， the Recall index improved by 2.41-6.91 percentage points， and the F1 value improved by 2.19-7.19 percentage points. In addition， the SHAP model yields the core factors affecting breast cancer， such as concave points_worst， perimeter_worst， and area_worst， which provide the principle support for doctors’ diagnosis. Reference | Related Articles | Metrics | Comments（0）

Select

SOC Estimation of Lithium Battery Based on Improved LSTM PAN Si-yuan, ZHANG Wei Computer and Modernization    2023, 0 (08): 25-30.   DOI: 10.3969/j.issn.1006-2475.2023.08.005 Abstract （124）      PDF（pc） （3576KB）（70）       Knowledge map    Save Absrtact: Aiming at the low accuracy of the state of charge（SOC） estimation of lithium batteries， a neural network model based on improved LSTM algorithm is proposed to obtain the mapping relationship between voltage and current input and SOC output. By extending the Kalman filter to filter the noise of the output estimate， the stability of the model is enhanced. In the process of neural network modeling， the improved particle swarm optimization algorithm is used to optimize the number of neurons， learning rate， step size and other super parameters， which further improves the efficiency and accuracy of lithium battery SOC estimation. Finally， the DST condition data in the university of Maryland CALCE dataset is used for model training， and the FUDS and US06 condition data-sets are used for comparative experiments on the improved LSTM algorithm， CNN-LSTM、GRU algorithm and CatBoost algorithm. The experimental results show that the improved LSTM estimation model has high stability and accuracy， which verifies the feasibility of the improved scheme. Related Articles | Metrics | Comments（0）

Select

An Improved Sparrow Search Algorithm Based on Multi-strategy LU Lei, HE Zhi-ming, HUANG Zhi-cheng Computer and Modernization    2023, 0 (10): 23-31.   DOI: 10.3969/j.issn.1006-2475.2023.10.004 Abstract （123）      PDF（pc） （2714KB）（70）       Knowledge map    Save  To address the problems that the population diversity of the sparrow search algorithm （SSA） decreases in the late iteration and easily falls into local optimum， a multi-strategy based improved sparrow search algorithm （MUSSA） is proposed. Firstly， MUSSA uses opposition-based learning and iterative strategy to enhance population diversity. According to the forgotten decline strategy， the number of populations using the reverse iteration strategy is gradually reduced， the loss of useless search is reduced， and the convergence speed of the algorithm is accelerated. Then， the adaptive weight spiral search strategy and reference frame mechanism are introduced to modify the discoverer update formula， further expand the search range of individuals and enhance the global search capability of the algorithm. Finally， direction factor and non-static selection strategy are introduced into follower renewal strategy to enhance local mining excavation. The simulation results of 13 benchmark test functions show that MUSSA has better optimization performance than SSA， HHO， WOA and AO. Reference | Related Articles | Metrics | Comments（0）

Select

Real-time Behavioral Safety Warning for Power Operators Based on Multi-source Data ZHANG Nan, LI Wen-jing, LIU Cai, XIE Ke, MA Shi-qian, XIAO Jun-hao, ZOU Feng Computer and Modernization    2023, 0 (10): 84-91.   DOI: 10.3969/j.issn.1006-2475.2023.10.013 Abstract （120）      PDF（pc） （2742KB）（84）       Knowledge map    Save  In order to reduce the occurrence of safety accidents and ensure the safety of power operators in the process of power grid construction， a behavior recognition model based on decision fusion of three dimensional residual convolutional neural network （R3D） models is proposed. First， the captured video dataset is subjected to data cleaning and enhancement； then， the corresponding R3D models are trained with the datasets collected from multiple angles； further， the multiple R3D models are fused at the decision level； finally， the possible violations or dangerous behaviors of power operators are warned in real-time by building a cloud platform. The experimental results show that the model has the advantages of high recognition accuracy and a low number of parameters， which proves that the behavior safety early warning method proposed in this paper can make early warning quickly and accurately and provide a safety guarantee for power grid construction. Reference | Related Articles | Metrics | Comments（0）

Select

Multi Path Planning Based on Constrained Clustering and Particle Swarm Optimization HAN Xue Computer and Modernization    2023, 0 (08): 7-11.   DOI: 10.3969/j.issn.1006-2475.2023.08.002 Abstract （119）      PDF（pc） （1639KB）（66）       Knowledge map    Save Abstract: In Large-scale logistics center ， if logistics management information system can be used normally， it is necessary to study the problem of vehicle routing in multi-distribution centers.We want to use as few vehicles as possible to complete the delivery of goods and minimize the total mileage.K-shortest paths in multi-center path planning has conducted in-depth research， the multi-path planning problem has been realized by using the traditional clustering algorithm.However， in the real multi-distribution-center vehicle routing planning， there are specific restrictions on the transportation capacity of transportation vehicles and the needs of users. We introduce constraint mechanism on the clustering algorithm to reduce the dimension of multi distribution center problem to single distribution center problem by clustering algorithm， and particle swarm optimization is introduced to solve the optimal solution of multi-path planning for single distribution center.The experiment proves the superiority of this method：the practice proves that the convergence speed of this method is at least n （number of distribution centers) times faster than that of the traditional particle swarm optimization algorithm， which provides a new solution for path planning. Related Articles | Metrics | Comments（0）

Select

Fault Diagnosis of Hydraulic Systems Based on CNN-BiLSTM LIU Fu-qi, ZHANG Da, SONG Jian-hua, WANG Hai-dong Computer and Modernization    2023, 0 (09): 10-19.   DOI: 10.3969/j.issn.1006-2475.2023.09.002 Abstract （119）      PDF（pc） （8241KB）（71）       Knowledge map    Save Aiming at the fault diagnosis problem of the main components in complex hydraulic system， a fault diagnosis model based on one-dimensional convolutional neural network （1D-CNN） and bidirectional long-term memory network （BiLSTM） is proposed to achieve multi-sensor information fusion and make fault diagnosis of piston pump and throttle valve. In the proposed model， the signals collected by various sensors are carried out data-stage fusion firstly， then the fault characteristics of the fusion signal are extracted by CNN and dimensionality reduction is performed， and then the forward and reverse data characteristics in the signal are learned by BiLSTM， finally the Softmax is used for classification， which realizes the diagnosis of piston pump and throttle valve fault. The experimental results show that the proposed method can automatically extract the fault characteristics in the signal and consider the positive and negative data characteristics contained in the signal. The diagnostic accuracy of the plunger pump can reach 96.3%， and the diagnostic accuracy of the throttle valve can reach 94.28%， which realizes the accurate and reliable diagnosis of the fault state of the plunger pump and the throttle valve. Reference | Related Articles | Metrics | Comments（0）

Select

Research Status and Prospect of Edge Computing in Smart Distribution Network HE Yu-peng, TAO Yong, WANG Bing-heng, ZHAO Ying-nan Computer and Modernization    2023, 0 (08): 87-92.   DOI: 10.3969/j.issn.1006-2475.2023.08.014 Abstract （118）      PDF（pc） （1132KB）（68）       Knowledge map    Save Abstract： With the rapid development of artificial intelligence technology and the Internet of Things， the distribution network is gradually becoming intelligent. However， massive data faces cloud computing face problems such as longer delay， network congestion， and privacy leakage. As a new computing paradigm， edge computing can handle network edge nodes and effectively solve the above problems， and is increasingly used in smart distribution networks. This paper reviews the edge computing technology of smart distribution networks in recent years. Firstly， it summarizes the characteristics of smart distribution network and the definition and architecture of edge computing in this application scenario. Secondly， it summarizes typical applications from different dimensions， including fault diagnosis and detection， data analysis， optimal scheduling， and data security and protection.. The research challenges in grid scenarios are summarized and prospected in three aspects， refinement management of data， modular sharing of resources and edge security maintenance. Related Articles | Metrics | Comments（0）

Select

GAN-based Adversarial Attacks on Face Recognition WANG Xin, XIAO Tao-rui Computer and Modernization    2023, 0 (10): 115-120.   DOI: 10.3969/j.issn.1006-2475.2023.10.017 Abstract （118）      PDF（pc） （1551KB）（75）       Knowledge map    Save  Face recognition is gradually becoming a monitoring tool which posed enormous threats to human privacy. For this reason， the paper proposes a semantic adversarial attack based on generative adversarial networks called SGAN-AA that modifies the significant facial features for images. It predicts the most significant attributes by using cosine similarity or probability score， and uses one or more facial features in white-box and black-box settings for impersonation and dodging attacks. The experimental results show that the method can generate diverse and realistic adversarial facial images while avoiding affecting human perception of facial recognition. The success rate of SGAN-AA's attack on black box models is 80.5%， which is 35.5 percentage points higher than common methods under impersonation attacks. Predicting the most significant attributes will improve the success rate of adversarial attacks in both white-box and black-box settings， and can enhance the transferability of the generated adversarial examples. Reference | Related Articles | Metrics | Comments（0）

Select

Cross-language Multi-label Sentiment Classification Based on Stacked Denoising AutoEncoder TANG Shi-qi, ZHOU Rui-ping, XIE Shi-bin, LIU Meng-chi, XIAO Wen, Computer and Modernization    2023, 0 (11): 6-12.   DOI: 10.3969/j.issn.1006-2475.2023.11.002 Abstract （118）      PDF（pc） （1392KB）（81）       Knowledge map    Save Abstract: The multi-label sentiment classification task aims to deal with the problem that an instance may be associated with multiple sentiment labels. Most existing multi-label sentiment classification models were designed based on complete data，and their performance and sentiment were easily affected by the incompleteness of data itself. To address this problem，a cross-language multi-label sentiment classification model based on stacked denoising autoencoder is proposed， and a loss function is introduced to compensate for the loss caused by training. In this model， the word vectors are denoised by the stacked denoising autoencoder to construct the low-dimensional features of the original data. This reduces the noise interference in feature space and provides effective feature representation for downstream tasks. In the multi-label sentiment classification experiment of SemEval2018 three language datasets （English， Arabic and Spanish）， the micro_F1 score， macro_F1 score and jaccard indexes of the model on the test set are all improved. Macro_F1 is improved by about 0.82， 1.45 and 1.83 percentage points， respectively. Reference | Related Articles | Metrics | Comments（0）

Select

Review and Discussion of Personalised News Recommendation Systems ZHAI Mei Computer and Modernization    2024, 0 (04): 12-20.   DOI: 10.3969/j.issn.1006-2475.2024.04.003 Abstract （117）      PDF（pc） （1534KB）（84）       Knowledge map    Save Abstract： With the rapid development of news media technology and the exponential growth of the number of online news， personalised news recommendation plays an extremely crucial role in order to solve the problem of online information overload. It learns users' browsing behaviour， interests and other information， and actively provides user with news of interest， thus improving user's reading experience. Personalised news recommendation has become a hot research and practical problem in the field of journalism and computer science， and experts in the industry have proposed various recommendation algorithms to improve the performance of recommendation systems. In this paper， we systematically describe the latest research status and progress of personalised news recommendation. firstly， we briefly introduce the architecture of news recommendation systems， and then we study the key recommendation algorithms and common evaluation metrics in news recommendation systems. Although personalised news recommendation brings a good experience to users， it also brings a lot of unknown effects to users. Unlike other news recommendation reviews， this paper also examines the impact of current news recommendation systems on user behaviour and the problems they face. Finally， the paper proposes research directions and future work on personalised news recommendation based on the current problems encountered. Reference | Related Articles | Metrics | Comments（0）

Select

HRNet Image Semantic Segmentation Algorithm Combined with Attention Mechanism YE Si-jia, WEI Yan, DU Han-yu, DENG Jin-zhi Computer and Modernization    2023, 0 (10): 65-69.   DOI: 10.3969/j.issn.1006-2475.2023.10.010 Abstract （116）      PDF（pc） （2303KB）（66）       Knowledge map    Save Abstract: The current mainstream semantic segmentation algorithms still have problems such as loss of small-sized objects and inaccurate segmentation. In response to these problems， this paper improves the HRNet network model and integrates the attention mechanism to generate more effective feature maps. To address the problem of insufficient detail of the feature map caused by the direct fusion of the low resolution images to the high-resolution images in the original model， a multi-level upsampling mechanism is proposed to make the fusion between images of different resolutions smoother to achieve better fusion results， and the depth separable convolution is used to reduce the parameters of the model. The model in this article maintains a high resolution of the image throughout the entire process. The spatial information of the feature map is improved， and the segmentation effect of small-sized objects is improved. The mIoU value on the PASCAL VOC2012 enhanced dataset reaches 80.87%， and the accuracy is improved by 1.54 percentage points compared with the original model. Reference | Related Articles | Metrics | Comments（0）

Select

Automatic Arrangement Method of Cloud Network Security Service Chain Based on SRv6 Technology WANG Hong-jie, XU Sheng-chao, YANG Bo, MAO Ming-yang, JIANG Jin-ling Computer and Modernization    2024, 0 (01): 1-5.   DOI: 10.3969/j.issn.1006-2475.2024.01.001 Abstract （101）      PDF（pc） （1156KB）（173）       Knowledge map    Save Abstract: To improve the resource utilization rate of cloud network data centers and save communication costs， a cloud network security service chain automatic orchestration method is designed based on SRv6 （Segment Route IPv6） technology. The method assists and guides network data packets to pass through the cloud network along the specified path， determines the specific forwarding path of the message， and reduces dependence on service nodes； establishes an objective function to minimize the total bandwidth， combines with various constraints to meet the security requirements of automatic orchestration； defines local behavior message， constructs automatic arrangement framework of security service chain， establishes security service policy， solves policy conflict and flow network scheduling problem， and achieves security arrangement of service chain. Experimental results show that the proposed method can effectively implement the automatic scheduling of cloud service chain， reduce the average total bandwidth consumption of CPU， improve the success rate of user requests， reduce the load of edge device in the cloud， and save communication costs. Reference | Related Articles | Metrics | Comments（0）

Select

Named Entity Recognition in Electronic Medical Record Based on BERT ZHENG Li-rui, XIAO Xiao-xia, ZOU Bei-ji, LIU Bin, ZHOU Zhan Computer and Modernization    2024, 0 (01): 87-91.   DOI: 10.3969/j.issn.1006-2475.2024.01.014 Abstract （99）      PDF（pc） （992KB）（265）       Knowledge map    Save Abstract：Electronic medical record is an important resource for the preservation， management and transmission of patients’medical records. It is also an important text record for doctors’ diagnosis and treatment of diseases. Through the electronic medical record named entity recognition （NER） technology， diagnosis and treatment information such as symptoms， diseases and drug names can be extracted from the electronic medical record efficiently and intelligently. It is helpful for structured electronic medical records to use machine learning and other technologies for diagnosis and treatment regularity mining. In order to efficiently identify named entities in electronic medical records， a named entity recognition method based on BERT and bidirectional long short-term memory network （BILSTM） with fusion adversarial training （FGM） is proposed， referred to as BERT-BILSTM-CRF-FGM （BBCF）. After preprocessing by correcting the Chinese electronic medical record corpus provided by the 2017 National Knowledge Graph and Semantic Computing Conference （CCKS2017）， the BERT-BILSTM-CRF-FGM model is used to recognize five types of entities in the corpus， with an average F1 score of 92.84%. Compared to the BERT model based on the inflated convolutional neural network （BERT-IDCNN-CRF） and the conditional random field model based on BILSTM （BILSTM-CRF）， the proposed method has higher F1 score and faster convergence speed， which can more efficiently structure electronic medical record text. Reference | Related Articles | Metrics | Comments（0）

Select

Garbage Classification and Detection Method Based on Improved YOLOX OUYANG Fei, WU Xu, XIANG Dong-sheng Computer and Modernization    2023, 0 (08): 68-73.   DOI: 10.3969/j.issn.1006-2475.2023.08.011 Abstract （99）      PDF（pc） （2467KB）（72）       Knowledge map    Save Abstract: Garbage classification and recycling can improve environmental pollution， protect residents’ living environment and ensure sustainable ecological development. However， traditional artificial garbage classification methods are inefficient and subjective. This paper proposes a garbage classification and detection method based on improved YOLOX to improve the efficiency and accuracy of garbage classification. By training YOLOX network on self-made garbage classification dataset， garbage detection and classification have been realized. In order to achieve better detection effect， ECA attention mechanism is introduced into the network to improve the information transmission ability between features. Improving the up sampling and down sampling times of the feature extraction network to improve the feature extraction ability of small targets. The classification and regression loss functions are improved to improve the learning ability of the network. The experimental results show that the mAP@0.75 of the improved YOLOX algorithm is 89.9%， which is 4 percentage points higher than that of the original algorithm， and the number of detected frames per second only decreases by 0.3. The detection accuracy is significantly improved without loss of performance. Related Articles | Metrics | Comments（0）

Select

Personnel Safety Warning System in Industrial Plant Based on Computer Vision GU Cheng-wei, DING Yong, LI Deng-hua Computer and Modernization    2023, 0 (09): 20-26.   DOI: 10.3969/j.issn.1006-2475.2023.09.003 Abstract （99）      PDF（pc） （2730KB）（66）       Knowledge map    Save In view of the frequent safety accidents of hoisting machinery in industrial plants， this paper proposes a personnel safety alert system in industrial plants based on computer vision， which uses a combination of computing platform and target detection algorithm to detect the personnel targets in the field operation monitoring video in real time and output corresponding control instructions. The target detection algorithm is based on YOLOv5 network， and the attention mechanism is embedded in the network structure. The space and channel based hybrid attention mechanism module is added to BottleneckCSP module， which can improve the accuracy of small target detection. In addition， a person tracking algorithm is introduced to modify and fuse the detection results， which can reduce the missed detection rate when the person is in the occlusion situation. The improved algorithm is tested in the self built dataset. Compared with the original YOLOv5 network， the improved algorithm is 3.414 percentage point higher on the mAP， and the detection speed can reach 40.3 FPS， which has a good detection effect. Finally， the algorithm model is deployed to the computing platform， and is built and tested on the scene. The test statistics showe that the detection accuracy of ordinary personnel and navigators is 94.4% and 95.1%， respectively， which has good detection performance and can stably perform corresponding automatic security alert operations. Reference | Related Articles | Metrics | Comments（0）

Select

Surface Anomaly Detection Algorithm of Flexible Plastic Packaging Based on Improved ConvNeXt NONG Hao-cheng, REN De-jun, REN Qiu-lin, LIU Peng-li, HUANG De-cheng Computer and Modernization    2023, 0 (08): 12-17.   DOI: 10.3969/j.issn.1006-2475.2023.08.003 Abstract （98）      PDF（pc） （1995KB）（62）       Knowledge map    Save Abstract： As for the artificial detection of flexible plastic packaging is slow and easily influenced by subjective factors which bring the problems such as error checking，as well as machine vision based on the deep learning only got a few of negative sample  which is difficult to obtain， the article proposed a ConvNeXt based asymmetirc dual network method to detect the outer surface of the tissue which is taken as the research object. Firstly， the method of machine vision based on threshold segmentation and image filtering is used to preprocess the image foreground extraction and correction， according to the situation of the industrial field images collected. Then， the anomaly detection network structure is constructed according to the characteristics of images. Finally， the preprocessed images were constructed as data sets to train and test the surface quality detection network of tissue. As a result， the experiment shows that the image-level AUROC is 99.75%， the pixel-level AUROC is 99.37%， and the detection time is 45 ms. The result meets the requirements of industrial real-time detection. Related Articles | Metrics | Comments（0）

Select

Collaborative Device-based Large-scale Offloading: A Bi-level Optimization Algorithm Fusing Divide-and-conquer and Greedy YAN Yang, ZHAN Zi-jun, CAO Shao-hua Computer and Modernization    2023, 0 (11): 13-21.   DOI: 10.3969/j.issn.1006-2475.2023.11.003 Abstract （98）      PDF（pc） （2607KB）（81）       Knowledge map    Save Abstract: With the rapid development of communication technology， the number of mobile devices is constantly increasing， which will also lead to frequent large-scale offloading scenarios. However， solving large-scale offloading problems in polynomial time remains a challenge. In this paper， we propose a bi-level optimization algorithm based on the cooperative computing network architecture， called DCGreedy， which fuses divide-and-conquer and greedy. This algorithm can efficiently solve the offloading strategy and resource allocation scheme of all tasks in polynomial time. It can effectively reduce the total energy consumption of the system while meeting all constraints. We evaluate the performance of DCGreedy based on the total number of tasks meeting deadlines， total system energy consumption， and algorithm runtime in a simulation scenario of at least 400 mobile devices. We conducted extensive experimental comparisons between DCGreedy and four other benchmark algorithms and found that in different scale offloading scenarios， the average total energy consumption of DCGreedy was 2.11% higher than the second ranked algorithm， while the algorithm’s running time was only 0.0049%. This fully confirms that DCGreedy effectively reduces the algorithm’s running time while optimizing system energy consumption. Reference | Related Articles | Metrics | Comments（0）

Select

Short-Term Natural Gas Load Forecasting Based on SARIMA Model SHAO Bi-lin, CHENG Wan-rong Computer and Modernization    2023, 0 (08): 54-59.   DOI: 10.3969/j.issn.1006-2475.2023.08.009 Abstract （94）      PDF（pc） （2381KB）（35）       Knowledge map    Save Abstract： Natural gas load forecasting plays a decisive role in residential life， commercial development and industrial production. And accurate short-term load forecasting can effectively quantify the uncertainty of natural gas load forecasting， which is critical for energy system operation and scheduling risk avoidance. The natural gas load affected by the seasonal effects will appear giant peak characteristics， the traditional point prediction model does not take into account the seasonal effects of natural gas， the accuracy of the prediction results is low. The SARIMA model can handle time series data with seasonal fluctuation trends and stochastic disturbances. Therefore， the SARIMA model is used to de-periodize the natural gas load as well as the first-order difference， capture the linear and seasonal features in the time series， determine the optimal parameter model based on the red pool information criterion and grid search， and proportionally divide the short-term interval forecast values. Taking the natural gas usage in Xi’an as an example， the results show that the SARIMA model used has a small error in the strong seasonal interval of the series and has a high accuracy when compared with the traditional model. Related Articles | Metrics | Comments（0）

Select

Pedestrian Detection Algorithm for Ship-borne Vehicles Based on YOLOX Combined#br# with DeepSort LIU Yu-shan, LIU Wei-kang, LIU Qing-hua, ZHE Tian-tian, WANG Jia-cheng Computer and Modernization    2023, 0 (08): 60-67.   DOI: 10.3969/j.issn.1006-2475.2023.08.010 Abstract （94）      PDF（pc） （1515KB）（55）       Knowledge map    Save Abstract：Aiming at the lack of real-time capture， detection and tracking of boarding vehicles and pedestrians in the current domestic ferry vehicle pedestrian control， this paper proposes a ship-borne vehicle and pedestrian detection method based on improved YOLOX. Firstly， the enhanced channel attention module is added to the three output heads of the enhanced feature extraction network of the original model to improve the feature extraction capability of the network for vehicles and pedestrians. Secondly， we use the improved ASPP module to replace the original SPP module. Among them， the improved ASPP module prunes the original module， and uses the addition of atrous convolution layers with different atrous convolution rates to solve the problem of local information loss of the original ASPP module. After the model is trained and verified with the validation set， it is combined with DeepSort for tracking detection. Compared with the original YOLOX algorithm， the average accuracy index （mAP） of the improved algorithm in this paper is increased by 3.3%， the accuracy rate is increased by 4.4%， and the test running speed on the GPU reaches 55 FPS. The experimental results show that the improved algorithm in this paper is suitable for real-time detection of vehicles and pedestrians in the ferry entrance environment. Related Articles | Metrics | Comments（0）

Select

Point Cloud Registration Algorithm Based on Combined Feature Points and#br# Principal Component Analysis#br# ZHANG Ya-wen, LIN Wen-zhong, HAN Xiao-dong Computer and Modernization    2023, 0 (09): 59-63.   DOI: 10.3969/j.issn.1006-2475.2023.09.009 Abstract （92）      PDF（pc） （1711KB）（43）       Knowledge map    Save Aiming at the problems of low accuracy， easy mismatching， and the descriptive error of single point features in the subsequent series of improved algorithms to point cloud shape， a point cloud registration algorithm based on point cloud combination feature point and principal component analysis is proposed. The intrinsic shape signatures is extracted from the point cloud， and the AC algorithm is used to extract the boundary points（BDRY） of the point cloud to form the combined feature points （ISS_BDRY）. The normal of the ISS_BDRY feature point is calculated and described by fast point feature histogram， and then the sampling consistency initial registration algorithm improved by principal component analysis SAC-IA is used to minimize the distance error between the main axes of the point cloud， thereby reducing the number of iterations in the point cloud fine registration process， and providing good pose for subsequent point cloud registration. In the fine registration stage， the iterative closest point registration algorithm introduced KD-Tree to accelerate search point cloud is used for registration. The experimental results show that compared with other single-point features， the registration accuracy of extracted combined feature points on Cat and Michael point clouds reaches 10-8 orders of magnitude， and the registration accuracy of the combined feature method is increased by 65.19% and 44.77%， respectively. Compared with ICP， NDT， Super 4PCS and other algorithms， the accuracy of the fine registration stage reaches 10-16 orders of magnitude， and it is almost completely coincide. Reference | Related Articles | Metrics | Comments（0）

Select

Behavior Recognition Method Based on FMCW Radar and ResNeSt-GRU MA Ze-yu, YE Ning, XU Kang, WANG Su, WANG Ru-chuan, Computer and Modernization    2023, 0 (11): 101-107.   DOI: 10.3969/j.issn.1006-2475.2023.11.016 Abstract （92）      PDF（pc） （2574KB）（100）       Knowledge map    Save Abstract： Aiming at the application of frequency modulated continuous wave radar in behavior recognition， a human behavior recognition system based on split attention residual neural network （ResNeSt） and gated neural unit （GRU） is proposed. The frequency modulated continuous wave （FMCW） radar is used to collect human behavior data. The fast Fourier transform algorithm （FFT） is used to extract the distance， velocity and angle dimension information of each frame of radar data， and then stitch them according to the time dimension into Range-Time Map （RTM）， Doppler-Time Map （DTM） and Angle-Time Map （ATM）. Finally， RTM， DTM and ATM are used as input samples， and the three-stream ResNeSt-GRU model is used to recognize different human behaviors. The experimental results show that the average recognition accuracy of the three-stream ResNeSt-GRU model for 8 behaviors reaches 98.92%， which is higher than the traditional deep learning model and the fusion deep learning model. In addition， the recognition accuracy rate using this model is 2.3% higher than that using a single-stream network after traditional feature fusion. Therefore， the system can effectively improve the recognition accuracy of the human behavior recognition system， and provide a new technology for the human behavior recognition. Reference | Related Articles | Metrics | Comments（0）

Select

An Environmental Target Recognition Method for Airport Special Vehicle Operation LIU Xu, ZHA Ke-ke Computer and Modernization    2023, 0 (08): 18-24.   DOI: 10.3969/j.issn.1006-2475.2023.08.004 Abstract （91）      PDF（pc） （2221KB）（65）       Knowledge map    Save Abstract: The autonomous and safe operation of airport special vehicles is essential to ensure the safety of the airfield area. At present， most airport special vehicle operations are mainly completed by driver’s operation and the visual command of controllers， in which such challenges as over-reliance on manpower and low autonomy. To improve its safety and autonomy， this paper presents a target recognition method for the airport environment based on the 3D point cloud segmentation. Firstly， a simulation-based approach is used to construct a point cloud dataset （Airfield Area of Airport Point Cloud Data，3A-PCD） of the airfield area environment. Secondly， based on PointNet++， a semantic segmentation network 3A-Net for large-scale point cloud data is designed， and a combined sampling point spatial encoding module and attentive pooling module are proposed to address the problem of traditional segmentation networks in terms of low segmentation accuracy and lack of ability to retain detailed features of objects. Finally， experiments were designed based on the 3A-PCD dataset， the ablation experiment result shows that the MIoU of the model increases by 6.0 percentage points with the addition of the spatial encoding module and by 3.9 percentage points with the addition of the AP module. 3A-Net achieves a 6.7 percentage points improvement in MIoU compared to the benchmark model PointNet++. In comparison with 6 existing advanced semantic segmentation models， the performance of the proposed model has been improved to varying degrees and is more suitable for target recognition in large outdoor scenes. Key words：target recognition； airport special vehicle； semantic segmentation； attention mechanism； 3D scene simulation Related Articles | Metrics | Comments（0）