Top Read Articles

    Published in last 1 year |  In last 2 years |  In last 3 years |  All
    Please wait a minute...
    For Selected: Toggle Thumbnails
    Road Pothole Detection Algorithm Based on Improved YOLOv5s
    BAI Rui, XU Yang, WANG Bin, ZHANG Wen-wen
    Computer and Modernization    2023, 0 (06): 69-75.   DOI: 10.3969/j.issn.1006-2475.2023.06.012
    Abstract545)      PDF(pc) (3457KB)(55)       Save
    Aiming at the problem that existing target detection algorithms are difficult to accurately detect road potholes and the detection speed is slow, a road pothole detection algorithm based on improved YOLOv5s is proposed. Firstly, CA (Coordinate attention) module is integrated into YOLOv5s backbone network, so that the model can capture not only cross-channel information, but also direction perception and position sensitive information, which is helpful for the model to locate and identify the detected object more accurately. Then, SoftPool is adopted in Spatial Pyramid Pool (SPP) module to improve the maximum pooling operation and retain more detailed characteristic information. In the feature fusion stage, Content-Aware ReAssembly of FEatures (CARAFE) is used to improve the up-sampling of multi-scale feature fusion and dynamically generate an adaptive kernel, which can gather context information in a large receptive field. Finally, Alpha-IoU is used to improve the loss function and improve the margin regression accuracy. Experimental results show that the average accuracy of the improved YOLOv5s algorithm is 4.6 percentage points higher than that of the original network, and the detection accuracy of the improved YOLOv5s algorithm is greatly improved compared with other mainstream algorithms such as SSD, Faster R-CNN, YOLOv3, YOLOv3-tiny and YOLOv4-tiny.
    Reference | Related Articles | Metrics | Comments0
    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
    Abstract349)      PDF(pc) (2112KB)(84)       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 | Comments0
    A Hybrid Brain Tumor Classfication Study Based on CBAM and EfficientNet with Improved Channel Attention
    HUA Xin-yu, QI Yun-song
    Computer and Modernization    2023, 0 (05): 1-7.  
    Abstract348)      PDF(pc) (1818KB)(52)       Save
    In order to further improve the accuracy and robustness of brain tumor image diagnosis, a novel hybrid brain tumor classification method based on CBAM(Convolutional Block Attention Module) and EfficientNet with improved channel attention mechanism (IC+IEffxNet) is proposed. The method is divided into 2 stages. In the first stage, the features will be extracted by CBAM model based on improved spatial attention mechanism. In the second stage, the sequence and exception (SE) block in EfficientNet architecture is replaced by the efficient channel attention (ECA) block, and the combined feature output of the first stage is used as the input of the second stage. Experiment shows the 4 classifications of glioma, meningioma, pituitary and normal images from the mixed brain tumor MRI dataset. The results show that the average classification accuracy is about 0.5~2 percentage points higher than the existing methods. The experimental results demonstrate the effectiveness of the method and provide a new reference for medical experts to accurately judge brain tumor.
    Reference | Related Articles | Metrics | Comments0
    A Remote Sensing Image Change Detection Model Based on CNN-Transformer Hybrid Structure
    XU Ye-tong, GENG Xin-zhe, ZHAO Wei-qiang, ZHANG Yue, NING Hai-long, LEI Tao
    Computer and Modernization    2023, 0 (07): 79-85.   DOI: 10.3969/j.issn.1006-2475.2023.07.014
    Abstract342)      PDF(pc) (2633KB)(66)       Save
    The emergence of convolutional neural network and Transformer model has made continuous progress in remote sensing image change detection technology, but at present, these two methods still have shortcomings. On the one hand, the convolutional neural network cannot model the global information of remote sensing images due to its local perception of convolution kernel. On the other hand, although Transformer can capture the global information of remote sensing images, it cannot model the details of image changes well, and its computational complexity increases quadrally with the resolution of images. In order to solve the above problems and obtain more robust change detection results, this paper proposes a CNN-Transformer Change Detection Network (CTCD-Net) based on convolutional neural network and Transformer hybrid structure. Firstly, CTCD-Net uses convolutional neural network and Transformer based on encoding and decoding structure in series to effectively encode local and global features of remote sensing images, so as to improve the feature learning ability of the network. Secondly, the cross-channel Transformer self-attention module (CSA) and attention feedforward network (A-FFN) are proposed to effectively reduce the computational complexity of Transformer. Full experiments on LEVIR-CD and CDD datasets show that the detection accuracy of CTCD-Net is significantly better than that of other mainstream methods.
    Related Articles | Metrics | Comments0
    Event Extraction Method Based on BERT-BiLSTM-Attention Hybrid Model
    WEI Xin, HE Xiao-hai, TENG Qi-zhi, QING Lin-bo, CHEN Hong-gang
    Computer and Modernization    2023, 0 (04): 26-31.  
    Abstract327)      PDF(pc) (1263KB)(74)       Save
    Event extraction is one of the basic tasks in the information extraction’s field, which is aims to extract structured information from unstructured text. The majority of the existing event extraction methods which are based on machine reading comprehension model directly detect and classify the input text trigger words, and to some extent ignore the prediction error caused by judging whether the input text is an event. Therefore, this paper proposes an event extraction method based on BERT-BiLSTM-Attention hybrid model. This method takes BERT-based machine reading comprehension model as the basic model, adopts multi-round question-and-answer method, and adds event classification detection module on the basis of existing machine reading comprehension model to reduce prediction error. BiLSTM model is combined with attention mechanism to form historical session information module to more effectively filter out important information and integrate it into a reading comprehension model. The event extraction experiments are conducted on ACE2005, and the results show that the accuracy, recall and F1 value are improved by 7.8 percentage points, 4.6 percentage points and 5.4 percentage points, respectively, compared with the basic model, which has certain advantages.
    Reference | Related Articles | Metrics | Comments0
    Fault Diagnosis of Pumping Unit Based on 1D-CNN-LSTM Attention Network
    WANG Lei, ZHANG Xiao-dong, DAI Huan
    Computer and Modernization    2023, 0 (04): 1-6.  
    Abstract304)      PDF(pc) (1482KB)(71)       Save
    Aiming at the problems of complex feature extraction, large amount of model parameters and low diagnostic efficiency in traditional fault diagnosis methods of pumping unit based on dynamometer diagram, this paper proposes a fault diagnosis method based on 1D-CNN-LSTM attention network. The dynamometer diagram is converted into a load displacement sequence as the network input, the one-dimensional convolutional neural network (1D-CNN) is used to extract local features of the sequence while reducing sequence length. Considering the temporal characteristics of the sequence, the long-short-term memory (LSTM) network is further used to extract temporal features of the sequence. In order to highlight the impact of key features, the attention mechanism is introduced to give higher attention weights to temporal features related to fault type. Finally, the weighted features are input into a fully connected layer, and the Softmax classifier is used to realize fault diagnosis. The experimental results show that the average accuracy, precision, recall and F1 value of the proposed method reach 99.13%, 99.35%, 99.17% and 99.25%, respectively, and the model size is only 98 kB. Compared with other methods based on feature engineering, it has higher diagnostic accuracy and generalization. Compared with other methods based on two-dimensional convolutional neural network (2D-CNN) model, it significantly reduces model parameters and training time, improves the efficiency of fault diagnosis.
    Reference | Related Articles | Metrics | Comments0
    Optimization Method of Hadoop File Archiving Based on LZO
    ZHANG Jun, SU Wen-hao
    Computer and Modernization    0, (): 1-6.   DOI: 10.3969/j.issn.1006-2475.2023.06.001
    Abstract290)      PDF(pc) (948KB)(42)       Save
    The distributed framework Hadoop is widely used in various fields of big data processing. However, more metadata information will be generated while a large number of small files are stored in Hadoop, which can lead to excessive usage of memory in NameNode and affect its ability to provide high performance and high concurrent access. Archiving and storing small files is an effective solution to this problem. At the same time, as data compression can effectively reduce the size of data storage space and network data transmission load, this paper proposes a Hadoop file archiving optimization method named LA (LZO-Archive)based on a real-time lossless compression algorithm LZO. In order to reduce the time of generating index files, LA incorporates LZO compression algorithm during the process of the index file generation stage on the basis of archiving and merging small files. Moreover, a file compression storage algorithm is designed in LA to compress and store data files and index files, which can effectively reduce the occupied disk space in DataNode and the occupied memory space in NameNode. This paper also elaborates the design and implementation of experimental method for LA. Experimental results show that compared with the original HDFS data storage method, the benchmark method of file archiving HAR and the comparison method LHF, the proposed method LA performs better in the aspects of file archiving time, memory usage in NameNode, disk space usage in DataNode, and file access time.
    Reference | Related Articles | Metrics | Comments0
    Bearing Fault Diagnosis Based on CWGAN-GP and CNN
    JIANG Lei, TANG Jian, YANG Chao-yue, LYU Ting-ting
    Computer and Modernization    2023, 0 (07): 1-6.   DOI: 10.3969/j.issn.1006-2475.2023.07.001
    Abstract282)      PDF(pc) (2107KB)(48)       Save
    Abstract: Aiming at the problem that the number of bearing fault samples is small and unbalanced in the actual work process, a bearing fault diagnosis method based on Conditional Wasserstein Generative Adversarial Network (CWGAN-GP) and Convolutional Neural Network (CNN) is proposed. First, a CWGAN-GP generative adversarial network is constructed by combining conditional generative adversarial network (CGAN) and gradient penalized Wasserstein distance-based generative adversarial network (WGAN-GP). Then, a small number of bearing fault data samples are input into CWGAN-GP, in order to obtain high-quality samples similar to the original samples. When the network reaches the Nash equilibrium, the generated samples and the original samples are mixed to generate a new sample set. Finally, the new sample set is input into the convolutional neural network to learn the sample features for fault diagnosis. The experimental results show that the diagnostic accuracy of the diagnostic method proposed in this paper exceeds 99%, which is higher than other diagnostic methods, effectively improving the diagnostic accuracy and enhancing its generalization ability.
    Reference | Related Articles | Metrics | Comments0
    Small Object Detection Method Based on Improved YOLOv5
    WANG Yi-cheng, ZHANG Guo-liang, ZHANG Zi-jie,
    Computer and Modernization    2023, 0 (05): 100-105.  
    Abstract251)      PDF(pc) (2253KB)(50)       Save
    In order to solve the problems of low detection accuracy and missing detection in traditional YOLOv5 object detection algorithm, a small object detection method based on improved YOLOv5 was proposed. Firstly, to make anchor box better adapt to small targets, IOU (interp over union) is used to replace the Euclidean distance formula originally used in the K-means clustering process to redefine the distance between anchor box and ground truth. Secondly, a maximum pooling of 3×3 kernel size is added to spatial pyarmid pooling (SPP) to improve the detection accuracy of small targets. Finally, a data set containing a variety of small object is designed to verify the algorithm performance. Experimental results show that the mean average precision (mAP) of the improved YOLOv5 algorithm reaches 76.92%, which is 3.56 percentage points higher than that of the classical YOLOV5 algorithm. The detection performence is improved and missed object can be detected.
    Reference | Related Articles | Metrics | Comments0
    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
    Abstract231)      PDF(pc) (1213KB)(90)       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 | Comments0
    Multi-robot Path Planning Based on Double Fuzzy Inference and Improved DWA Algorithm
    WANG Zi-wei
    Computer and Modernization    2023, 0 (04): 20-25.  
    Abstract225)      PDF(pc) (2420KB)(51)       Save
    Aiming at the difficulties—balance the performance indexes such as reachability and safety with time in unknown complex scene—faced by existing multi-robot system path planning methods, an improved DWA (dynamic window approach) algorithm based on two-layer fuzzy inference is proposed. First, the linear velocity fuzzy controller and the steering angle fuzzy controller output the base pose to ensure the flexibility and safety of the robot path-planning process. Then, comparing with the traditional DWA algorithm, the obstacle distance evaluation function is improved and the danger zone-related evaluation function is also incorporated to achieve multi-robot collision avoidance. Also, the robustness and global performance are improved by extending the evaluation function and the weight parameters. Finally, the two-layer fuzzy inference is fused with the improved DWA algorithm, so the two-layer fuzzy controller is used to determine the approximate speed and direction, based on which the precise speed and steering angle are output using the improved DWA. Simulation experiments show that the proposed algorithm generates smoother trajectories and improves the operational efficiency and safety of multi-robot path planning.
    Reference | Related Articles | Metrics | Comments0
    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
    Abstract218)      PDF(pc) (2224KB)(76)       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 | Comments0
    Survey on Blockchain Security Protection
    ZHA Kai-jin, WANG Zhi-bo, HE Yue-shun, XU Hong-zhen
    Computer and Modernization    2023, 0 (06): 110-117.   DOI: 10.3969/j.issn.1006-2475.2023.06.018
    Abstract215)      PDF(pc) (1274KB)(53)       Save
    Blockchain technology, as one of the most popular technologies at present, has huge application value. At the same time, it is widely used in many key fields due to the high support of the country. The many characteristics of blockchain technology determine its application advantages in data sharing, digital storage, information tracing, security guarantee, etc., and at the same time, it also brings many security risks. Because of this, this article summarizes the content and related research progress of the blockchain infrastructure, security threats, and privacy protection schemes by studying on high-quality literature on blockchain security protection related research at home and abroad. Aiming to the development status of blockchain privacy protection technology, from the two aspects of encryption technology improvement and privacy protection technology fusion research,we analyze its impact on the development of blockchain, hoping to provide reference for blockchain security protection research.
    Reference | Related Articles | Metrics | Comments0
    Modeling Approach of Multi-level and Multi-resolution Grid Model for Strategy Campaign Wargame
    LI Hai-yan, WU Da-yu, LIU Qiang
    Computer and Modernization    2023, 0 (06): 27-32.   DOI: 10.3969/j.issn.1006-2475.2023.06.005
    Abstract210)      PDF(pc) (1273KB)(42)       Save
    The grid model is the core component of strategy campaign wargame environment model and influences other behavior models. Building multi-level and multi-resolution grid models is the basis of extending strategy campaign wargame, supporting fine-grained environment models and key tactical operations. In this paper, the modeling approach and models of multi-level and multi-resolution equal longitude and latitude division for strategy campaign wargame are put forward based on the military requirements and technical requirements. Firstly, grid map is divided by using the equal latitude and longitude quadrangle. Then the coordinate system cluster of equal latitude and longitude grid and related elements are defined. Finally, an example is given. By comparing with single-resolution hexagon grid modeling approach, the advantages are analyzed. The modeling approach can support multiple-resolution of strategy campaign wargame, and meet the requirements of running efficiency and environment model for strategy campaign wargame.
    Reference | Related Articles | Metrics | Comments0
    Asymmetric Deep Supervised Hashing with Attention Mechanism
    WANG Xin-yi, YIN Si-qing, HONG Jun
    Computer and Modernization    2023, 0 (05): 26-31.  
    Abstract199)      PDF(pc) (5364KB)(55)       Save
    With the advent of the era of big data, the information data on the Internet is growing exponentially. Among these data, image resource accounts for a very large proportion, so how to carry out accurate and efficient image retrieval from massive images has become one of the important research topics today. At present, there are some problems in large-scale image retrieval, such as poor retrieval performance and low accuracy due to the inability to effectively focus on the key areas of the image. Based on the above shortcomings, an asymmetric deep hash algorithm that integrates the attention mechanism is proposed, which is modified based on convolutional neural network. The existing mixed attention mechanism guided by semantic features is improved and embedded into the network, so that the hash network can analyze the global semantic information and local semantic information together. At the same time, a new quantization function is designed to reduce quantization error, so as to enhance the feature expression ability of hash coding. This method is compared with other hashing methods on the CIFAR-10 and NUS-WIDE datasets with evaluation standard mAP. The results show that the proposed network model can combine global and local spatial features well, and improve the image retrieval performance.
    Reference | Related Articles | Metrics | Comments0
    Pseudo-color Enhancement Method for Quantum Images Based on IBM Qiskit
    LIU Zhi-fei, ZHU Shang-chao, WEI Zhan-hong, ZANG Yi-ming, SUN Wen-tao, HU Guan-shi
    Computer and Modernization    2023, 0 (04): 47-55.  
    Abstract197)      PDF(pc) (4493KB)(69)       Save
    Aiming at the problem of quantum image enhancement, a pseudo-color enhancement method for quantum images based on rainbow coding is proposed. Firstly, the NEQR (Novel Enhanced Quantum Representation) model is used to represent grayscale images, then the quantum circuit of the RGB three-channel color conversion module is designed and optimized, and finally the QRMW (Quantum Representation of Multi Wavelength Images) model is used to represent pseudo-color images. In order to verify the effectiveness of the proposed method, 2×2 and 32×32 NEQR grayscale images are prepared on the IBM quantum computing framework Qiskit, and QRMW pseudo-color images of corresponding sizes are generated by measuring the collapse of the quantum circuit. The experimental results show that, compared with the classical and existing quantum image pseudo-color enhancement methods, this method only requires 958 quantum fundamental gates when processing images with a size of 2n×2n and a color depth of 2q. The time complexity is constant-level O(1), and the space complexity is O(2n+2q+3), which significantly reduces the quantum cost, and the information entropy and sharpness indicators of the processed image are good.
    Reference | Related Articles | Metrics | Comments0
    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
    Abstract196)      PDF(pc) (1192KB)(71)       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 | Comments0
    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
    Abstract193)      PDF(pc) (5630KB)(74)       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 | Comments0
    Image Caption Generation Method Based on Channel Attention and Transformer
    LIU Jing, CHEN Jin-guang
    Computer and Modernization    2023, 0 (05): 8-12.  
    Abstract192)      PDF(pc) (10305KB)(50)       Save
    Image caption generation refers to translating an image into a sentence by computer. Aiming at the problems of existing image caption generation tasks, such as insufficient use of local and global features of images and high time complexity, this paper proposes a hybrid structure image caption generation model based on Convolution Neural Networks (CNN) and Transformer. Considering the spatial and channel characteristics of the convolutional network, firstly, the light-weight and high-precision attention ECA is fused with the convolutional network CNN to form an attention residual block, which is used to extract visual features from the input image. Then the features are input into the sequence model Transformer. At the encoder, use self-attention learning to obtain the participating visual representations. At the language decoder,capture the fine-grained information in the caption and learn the interaction between the caption sequences. The model was validated on MSCOCO dataset, and the BLEU-1, BLEU-3, BLEU-4, Meteor, and CIDEr metrics were improved by 0.3, 0.5, 0.7, 0.4, and 1.6 percentage points, respectively.
    Reference | Related Articles | Metrics | Comments0
    Improved YOLOv5s for Small Vehicle Object Detection on Remote Sensing Image
    QIU Di-fa, YU Shu-fang, LIU Jin-hui, BI Meng-zhao
    Computer and Modernization    2023, 0 (05): 122-126.  
    Abstract190)      PDF(pc) (1941KB)(38)       Save
    Because of its good detection effect and low computational complexity, YOLOv5s is widely used in various target detection tasks. However, its large downsampling stride makes it difficult to obtain satisfactory results for small-sized vehicle detection in satellite remote sensing images. In order to improve the performance of small target detection, on the basis of YOLOv5s, the strategy of reducing the downsampling stride is adopted to protect the texture and geometric features of small targets in the vehicle, and the attention mechanism module is inserted in front of the detection head to suppress the interference of complex background. The tested results on the autonomous data set with a resolution of 0.5 m/pixel show that the AP, recall and precision of SA-YOLOv5s for vehicle target detection reached 94.1%, 99% and 87.3% respectively, which were 16.4, 6 and 5 percentage points higher than YOLOv5s, and showed good detection performance.
    Reference | Related Articles | Metrics | Comments0
    Natural Gas Load Forecasting Based on FCGA-LSTM and Transfer Learning
    ZHANG Zhi-xia, XIE Bao-qiang
    Computer and Modernization    2023, 0 (07): 7-12.   DOI: 10.3969/j.issn.1006-2475.2023.07.002
    Abstract187)      PDF(pc) (1521KB)(47)       Save
    Abstract:High precision natural gas load forecasting is of great significance to the smooth and efficient operation of natural gas pipeline network. Most of the existing natural gas load forecasting methods are based on the condition of sufficient historical data, and there is little research on the problem of natural gas load forecasting in areas lacking historical data. To solve these problems, a short-term natural gas load forecasting method based on long and short-term memory(LSTM)neural network optimized by Fuzzy Coded Genetic Algorithm (FCGA) and transfer learning is proposed. First, the source domain and the target domain are selected, and the FCGA-LSTM prediction model is constructed by using a large amount of historical load data in the source domain. After model training and testing, the source domain model is moved to the target domain lacking data as a whole, and then a small amount of data in the target domain is used to fine tune and retrain the model. Finally, the target domain load prediction model is obtained. Taking a new residential area in Xi’an as an example, the results show that the prediction accuracy of the prediction method based on FCGA-LSTM and transfer learning is improved by 15.6 percentage points and 35.2 percentage points respectively compared with the combination method of LSTM and transfer learning, LSTM under non transfer learning, which proves the effectiveness of the model. The proposed method has certain guiding significance for the prediction of natural gas load in new urban areas lacking historical data.
    Reference | Related Articles | Metrics | Comments0
    Automatic Classification Method of CNC Machine Tool Fault Text Based on CNN-BiLSTM
    XU Ya-xin, HE Ze-en, XU Xu-kan
    Computer and Modernization    2023, 0 (04): 7-14.  
    Abstract181)      PDF(pc) (1685KB)(52)       Save
    Small and medium-sized CNC machine tool firms have accumulated a large amount of fault maintenance data recorded in manual text during operation and maintenance. In order to accomplish efficient and accurate classification and help maintenance personnel carry out their work efficiently, this paper proposes a fault text classification and prediction approach based on convolution neural network and bi-directional long-short-term memory network. Firstly, the pre-processing is completed by creating a professional feature word database, and Word2Vec is used to train the word vector. Secondly, after the CNN layer extracts local features from the text vector, context features are extracted from the forward and backward LSTM. After the feature fusion and weighting of CNN and BiLSTM layers in the full connected layer, the full connected layer finds the output with the highest probability as the prediction result through the Softmax activation function, and presents the prediction accuracy of each category with the confusion matrix. Based on the fault data of an enterprise in the Yangtze River Delta, this paper makes an experimental analysis, and compares it with a single CNN and BiLSTM model. The experimental results indicate that the prediction accuracy of the new method is up to 94%, the average accuracy is increased by 11 percentage points, and the P value, R value and F value are all up to 95%, which can be used as an effective method in the field of small data volume fault text classification.
    Reference | Related Articles | Metrics | Comments0
    Vehicle Detection of Remote Sensing Images Based on Improved YOLOv5 Algorithm
    ZHU Li-qing, LI Xiang,
    Computer and Modernization    2023, 0 (05): 117-121.  
    Abstract179)      PDF(pc) (2658KB)(31)       Save
    An improved model based on YOLOv5s is proposed for the problem of target miss detection in remote sensing images with complex targets in the background and blurred imaging due to small vehicles. A new backbone network structure is designed for the improved model: RepVGG network is selected for the backbone feature extraction of the improved model, while an attention mechanism, CoordAttention, is added to the backbone network to improve the perception capability of the model for small targets. Multi-scale feature fusion is added to improve the detection accuracy of the improved model for small targets, and the loss function of border regression is chosen to use DIoU to help the improved model achieve more accurate localization. After experiments, it is demonstrated that the improved YOLOv5 model improves the detection accuracy by 5.3 percentage points for target detection in remote sensing images compared to the original model in small target vehicles, and improves the mAP by 16.88 percentage points compared to Faster R-CNN. The improved model can have a larger detection accuracy improvement compared with the mainstream detection algorithms, and has a better detection accuracy than the original YOLOv5s model for small vehicle detection in remote sensing images.
    Reference | Related Articles | Metrics | Comments0
    Intelligent Traffic Light Control System Based on Machine Vision
    GUO Zhi-dong, LIU Jun, YANG Fan, ZHOU Xin, CHEN Liang-liang
    Computer and Modernization    2023, 0 (04): 101-105.  
    Abstract163)      PDF(pc) (1392KB)(41)       Save
    In order to complete the real-time traffic light command task accurately, flexibly and efficiently, machine vision is introduced to design a set of intelligent traffic light control system based on machine vision to meet the needs of intelligent urban road traffic lights at this stage. The system uses OV2640 camera to take pictures of the road conditions, collect image information, and store the collected data. Then, based on the series of algorithms YOLO (You Only Look Once) in visual processing, it carries out real-time recognition of the road conditions in the data, and transmits the recognition results to the timing module. Finally, it obtains the most timely data and processes it on the traffic lights in time, and has achieved good results in the simulation, further promoting the application of intelligent transportation system.
    Reference | Related Articles | Metrics | Comments0
    Traffic Light Control Optimization Based On D3QN
    ZHANG Guo-you, SONG Shi-feng
    Computer and Modernization    2023, 0 (07): 30-35.   DOI: 10.3969/j.issn.1006-2475.2023.07.006
    Abstract156)      PDF(pc) (1524KB)(39)       Save
    Traffic lights play a vital role in controlling traffic at intersections. At present, the traffic lights at urban intersections mostly adopt the control strategy of fixed timing and fixed phase transformation, which is difficult to meet different traffic flow conditions. It has become one of the research hotspots in the field of intelligent transportation to design a control scheme that can adjust the traffic light transformation in real time according to the traffic flow at the intersection. However, the traffic flow at urban intersections is dynamic, so it is difficult to study it directly. In order to design an appropriate traffic light dynamic control scheme, the deep strong learning technology is introduced. The intersection traffic light control problem is abstracted into a reinforcement learning model, which is solved by D3QN algorithm. On this basis, considering the vehicles in different states, the state input and reward function are improved. Finally, the simulation experiments under different traffic flows are carried out on the traffic simulator SUMO. The experimental results show that after the model training becomes stable, the average queue length of the D3QN algorithm with improved reward function and state input is significantly improved compared with the traditional fixed control strategy and adaptive control strategy under three traffic flows, and the control effect is better then DQN algorithms and DDQN algorithms.
    Related Articles | Metrics | Comments0
    A Temporal Convolutional Knowledge Tracking Model Based on#br# Multiple Feature Extraction#br#
    XIE Shi-bin, LIU Meng-chi, TANG Shi-qi, ZHOU Rui-ping,
    Computer and Modernization    2023, 0 (07): 25-29.   DOI: 10.3969/j.issn.1006-2475.2023.07.005
    Abstract153)      PDF(pc) (2268KB)(47)       Save
    Knowledge tracing (KT) is a key technology in the field of educational data mining. It uses students’ historical learning records to predict students’ next answer performance. Aiming at the problem that the deep knowledge tracking model based on time convolution network (TCN) only uses students’ answer sequences and answer results, and ignores other behavior characteristics of students, a deep knowledge tracking model based on multi feature extraction (TKT-PCA) is proposed. The model uses principal component analysis (PCA) method to automatically extract hidden features in a variety of students’ answer behavior and learn their representation. It not only reduces the feature dimension and redundant information, but also fully evaluates students’ knowledge mastery. The experimental results show that the TKT-PCA has the better prediction performance compared with other knowledge tracking baseline models.
    Key words: deep learning; knowledge tracking; temporal convolution network; educational data mining; intelligent education
    Related Articles | Metrics | Comments0
    Overview of State Machine Inference Technology for Unknown Protocols
    SHENG Jia-jie, NIU Sheng-jie, CHENG Yang, FANG Wei-qing, ZHANG Yu-jie, LI Peng, HU Su-jun
    Computer and Modernization    2023, 0 (05): 58-67.  
    Abstract152)      PDF(pc) (2194KB)(35)       Save
    Protocol reverse engineering (PRE) describes the behavioral logic of the protocol, which is generally divided into 2 steps: protocol format extraction and state machine construction. These two steps are both interrelated and independent. PRE has important significance in the field of network security. In this paper, we have comprehensively sort out the relevant reference of protocol state machine inference. The research status and development trend of protocol state machine reasoning are summarized and analyzed. Firstly, we introduce the formal definition and basic principles of PRE and discuss the specific requirements of the main fields. Secondly, we analyze the state machine inference methods and divide them into three patterns: clustering method, state-related method, and polling state entity. Then we compare the inverse ability and time efficiency of the algorithms from different perspectives. Finally, the development trend of protocol state machine reasoning is prospected.
    Reference | Related Articles | Metrics | Comments0
    Tibetan Named Entity Recognition Based on Small Sample Learning
    YU Tao, ZHANG Ying, Yong Tso,
    Computer and Modernization    2023, 0 (05): 13-19.  
    Abstract151)      PDF(pc) (1601KB)(28)       Save
    The task of Tibetan named entity recognition is to identify the names of people, places and organizations in the text. This paper proposed a Tibetan named entity recognition method based on small sample learning. In the training process, the feature fusion of entity location information, word segmentation information and Tibetan syllable information in the form of dimensional splicing could better represent the boundary information of Tibetan long entities. Ablation experiments were designed to explore the effect of different feature information on model performance. The experimental results show that our method is effective, and the F1 value is improved by 22.22~38 percentage points compared with the baseline experiment.
    Reference | Related Articles | Metrics | Comments0
    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
    Abstract151)      PDF(pc) (1332KB)(43)       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 | Comments0
    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
    Abstract148)      PDF(pc) (8512KB)(45)       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 | Comments0
    Opinion Leaders Mining Method Based on Improved Hits Algorithm
    WANG Liu, ZHU Yi-xin, HAN Li-ying
    Computer and Modernization    2023, 0 (06): 39-42.   DOI: 10.3969/j.issn.1006-2475.2023.06.007
    Abstract148)      PDF(pc) (1065KB)(35)       Save
    Microblog has gradually become an important carrier of public opinion communication. The opinion leaders in online public opinion play a driving role in the process of public opinion communication. It is necessary to explore the opinion leaders in microblog for the management of social network public opinion. Considering microblog users' behaviors such as forwarding comments in the network, a two-layer network of microblog users' “forwarding comments”is constructed. By introducing the influence contribution factor and weight factor of users' interaction behavior to mine users' influence, a microblog users' influence evaluation algorithm based on Hits improved algorithm is proposed. The experimental results show that the F-score comprehensive index score of this model is better than PageRank algorithm and Hits algorithm. It can more accurately identify opinion leaders in microblogging community topics, effectively calculate the actual influence of microblogging users, and more accurately and effectively identify opinion leaders in a certain topic of microblogging community, which can provide reference for research on opinion leaders mining in social networks.
    Reference | Related Articles | Metrics | Comments0
    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
    Abstract145)      PDF(pc) (5102KB)(62)       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 | Comments0
    Alzheimer’s Disease Image Classification Based on Improved EfficientNet
    ZHU Jian-bo, GE Ming-feng, DONG Wen-fei
    Computer and Modernization    2023, 0 (06): 56-61.   DOI: 10.3969/j.issn.1006-2475.2023.06.010
    Abstract144)      PDF(pc) (2105KB)(43)       Save
    To improve the effectiveness of the convolutional neural network for Alzheimer’s disease MRI image classification, a convolutional neural network FAMENET is proposed, which integrates an adaptive attention mechanism and data enhancement technique to alleviate data imbalance by introducing a data augmentation technique and Focal Loss loss function. The network is reconfigured to reduce the number of model parameters and the computational effort of the network while maintaining accuracy. The adaptive attention mechanism is introduced to solve the information loss problem caused by the downsampling of input images for feature extraction. In a large number of comparative experiments on public datasets, the classification accuracy of FAMENET reaches 79.95% and the AUC value reaches 82.54%. The designed ablation experiments also fully demonstrate the effectiveness of the proposed modules and networks.
    Reference | Related Articles | Metrics | Comments0
    An Early Diagnosis Method of COVID-19 Infection Based on ResNeXt and Improved nnU-Net
    XU Hao, TIAN Zhen-yu, LI Chao-fan, CUI Xin-xin, YANG Jian-lan
    Computer and Modernization    2023, 0 (06): 21-26.   DOI: 10.3969/j.issn.1006-2475.2023.06.004
    Abstract144)      PDF(pc) (3376KB)(33)       Save
    The early infection of novel coronavirus pneumonia is characterized by increased lung turbidity and density. In order to solve the problem of difficulty in diagnosing and locating lung lesions in early patients with computed tomography, an experimental protocol for the diagnosis of COVID-19 (Corona Virus Disease 2019) with lung lesion segmentation by ResNeXt and a modified nnU-Net (no-new-Net) is proposed. The mean accuracy of ResNeXt model classification is 0.8554, the AUC area is 0.8951, the Precision is 0.8321, the F1 score is 0.8132, and the mean Dice coefficient of improved nnU-Net model lesion segmentation reaches 0.7663, which is a combined improvement of 16.4% compared with other models segmentation ability. The experimental results show that this scheme can enhance the ability to extract infection features from the early lung CT images of new crowns, and achieve efficient disease typing and accurate lesion segmentation.
    Reference | Related Articles | Metrics | Comments0
    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
    Abstract144)      PDF(pc) (3631KB)(80)       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 | Comments0
    Lightweight Speech Emotion Recognition for Data Enhancement
    CUI Chen-lu, CUI Lin,
    Computer and Modernization    2023, 0 (04): 83-89.  
    Abstract140)      PDF(pc) (2978KB)(36)       Save
    The use of deep learning for speech emotion recognition requires a large amount of training data. In this paper, the original speech is enhanced by adding Gaussian white noise and shifting the waveform to generate new speech signals in the preprocessing stage, which not only improves the recognition accuracy but also enhances the robustness of the model, given the shortage of existing speech emotion databases and the defects of overfitting caused by the small amount of data. At the same time, due to the excessive amount of parameters of the common convolutional neural network, a lightweight model is proposed, which consists of separable convolutional and gated recurrent units. Firstly, MFCC features are extracted from the original speech as the input of the model, and secondly, separable convolutional is used to extract the spatial information of speech, and gated recurrent units extract the temporal information of speech so that the temporal and spatial information can be used to characterize the speech emotion at the same time. It can make the prediction results more accurate. Finally, a fully connected layer with softmax is fed to complete the sentiment classification. The experimental results show that the model in this paper can not only obtain higher accuracy but also compress the model by about 50% compared with the benchmark model.
    Reference | Related Articles | Metrics | Comments0
    Intrusion Detection Method Based on Particle Swarm Optimization Combined with LightGBM
    PAN Yu-qing, ZHANG Su-ning, FENG Ren-jun, JING Dong-sheng
    Computer and Modernization    2023, 0 (04): 123-126.  
    Abstract139)      PDF(pc) (1170KB)(47)       Save
    With the development of the Internet, people enjoy the many conveniences it brings, but also face many threats, such as worms and Trojan horses. To defend against these malicious attacks, intrusion detection systems have been created. By detecting anomalies in the current network, intrusion detection systems can effectively detect attacks and take countermeasures. However, the accuracy of traditional machine learning algorithms in intrusion detection models is not high. Based on this, this paper proposes an intrusion detection model based on particle swarm optimization and LightGBM, specifically, an intrusion detection model is constructed by using the LightGBM method and a particle swarm algorithm is used to optimize the parameters of LightGBM. Experiments show that the method proposed in this paper can effectively improve the accuracy of the model, with 98.61% of accuracy, 98.25% of precision, 99.17% of recall rate and 98.70% of F1 score.
    Reference | Related Articles | Metrics | Comments0
    Nonlinear Process Fault Detection Based on KPCA and SSA Optimized SVM
    SHEN Zhi, LI Yuan
    Computer and Modernization    0, (): 15-20.   DOI: 10.3969/j.issn.1006-2475.2023.06.003
    Abstract139)      PDF(pc) (1185KB)(36)       Save
    To solve the problem of high characteristic dimension of nonlinear data generated by industrial process, a process fault detection algorithm based on Kernel Principal Component Analysis (KPCA) and Sparrow Search Algorithm (SSA) which is used to optimize the parameters of Support Vector Machine is proposed. Firstly, KPCA algorithm is used to extract linear and nonlinear features of industrial data. Secondly, the data after feature extraction is used as training samples to establish a classification SVM model, and SSA algorithm is used to optimize the kernel parameter and penalty factor of SVM. Finally, the optimized SVM model is applied to test samples for fault detection. In this paper, in order to verify the classification effect of the proposed algorithm, KPCA-SSA-SVM is compared with SVM, KPCA-GA-SVM (Genetic Algorithm, GA) by using a set of nonlinear numerical examples and Tennessee Eastman chemical process data, and the efficiency and superiority of the proposed algorithm is verified.
    Reference | Related Articles | Metrics | Comments0
    Improved Solar Cell Defect Detection Algorithm Based on YOLOv5s
    LUO Wei, LIU Si-yuan, XU Jian-xiang, DONG Tian-pei
    Computer and Modernization    2023, 0 (07): 119-126.   DOI: 10.3969/j.issn.1006-2475.2023.07.020
    Abstract138)      PDF(pc) (2142KB)(46)       Save
     In the process of manufacturing solar cells, due to the imperfect manufacturing processes and operational failure of humans, the defects, such as broken cell, crack, finger failure and silicon material missing might be found in the solar cells. A solar cell defect detection model based on YOLOv5s, namely YOLOv5s_CG, is proposed to improve the precision of the solar cell defects detection. The algorithm introduces convolutional attention mechanism (CBAM) blocks in different positions of the backbone network and feature fusion layer. The attention mechanism of the backbone network focuses on the global information, and the attention mechanism of the feature fusion layer focuses on the local information. At the same time, it enhances the features in both spatial and channel dimensions and uses the GIOU loss function to evaluate the detection effect of the algorithm. The proposed method is tested on the open source solar cell dataset which is re-labeled by the authors. The experimental results show that the overall mean average precision (mAP) of the YOLOv5s-CG algorithm reaches 75.1%. Compared with the algorithm of YOLOv5s, various types of defect detection accuracy have been improved, among which the accuracy of crack and silicon material missing has increased by 0.036 and 0.033 respectively, and the average accuracy (mAP) of all classes has increased by 0.026. Compared with the mainstream target detection algorithm of SSD, the overall mean average precision (mAP) has improved by 0.123. The algorithm can accurately detect the defects of solar cells, which could provide a better defects detection algorithm for real solar cell production.
    Related Articles | Metrics | Comments0
    Trend Prediction of Infectious Diseases Based on Logistic-GF-SEIR Model
    WU Le, CHEN Gang, LI Zhu
    Computer and Modernization    2023, 0 (05): 20-25.  
    Abstract137)      PDF(pc) (14136KB)(28)       Save
    In order to improve the prediction accuracy of the epidemic trend of new infectious diseases, this paper improves the traditional SEIR model and proposes the Logistic-GF-SEIR model. Firstly, based on historical data, the Logistic model is used to fit the cumulative rehabilitation, and the daily rehabilitation rate, infection rate and contact rate are inverted. Secondly, Gaussian model and Logistic model are used to fit the optimal time-varying parameters. Finally, the initial value of the model is initialized to predict the trend of epidemic population. Taking the epidemic development trend of Wuhan and Japan in the early stage of COVID-19 outbreak as an example, the simulation test is carried out and compared with Logistic, SEIR, ARIMA, BP neural network and other prediction models. The results show that the fitting and prediction performance of the Logistic-GF-SEIR model is better than other models in the prediction of the epidemic situation in Wuhan, and the root mean square error is better than other models in the prediction of the epidemic situation in Japan, which verifies the feasibility, effectiveness and robustness of the proposed model. It can provide a basis for China to formulate prevention and control policies for similar infectious diseases.
    Reference | Related Articles | Metrics | Comments0