Most Download articles

    Published in last 1 year| In last 2 years| In last 3 years| All| Most Downloaded in Recent Month | Most Downloaded in Recent Year|

    Published in last 1 year
    Please wait a minute...
    For Selected: Toggle Thumbnails
    Research Progress of Transformer Based on Computer Vision
    LIU Wenting, LU Xinming
    Computer Engineering and Applications    2022, 58 (6): 1-16.   DOI: 10.3778/j.issn.1002-8331.2106-0442
    Abstract629)      PDF(pc) (1089KB)(533)       Save
    Transformer is a deep neural network based on the self-attention mechanism and parallel processing data. In recent years, Transformer-based models have emerged as an important area of research for computer vision tasks. Aiming at the current blanks in domestic review articles based on Transformer, this paper covers its application in computer vision. This paper reviews the basic principles of the Transformer model, mainly focuses on the application of seven visual tasks such as image classification, object detection and segmentation, and analyzes Transformer-based models with significant effects. Finally, this paper summarizes the challenges and future development trends of the Transformer model in computer vision.
    Reference | Related Articles | Metrics
    YOLOv5 Helmet Wear Detection Method with Introduction of Attention Mechanism
    WANG Lingmin, DUAN Jun, XIN Liwei
    Computer Engineering and Applications    2022, 58 (9): 303-312.   DOI: 10.3778/j.issn.1002-8331.2112-0242
    Abstract671)      PDF(pc) (1381KB)(498)       Save
    For high-risk industries such as steel manufacturing, coal mining and construction industries, wearing helmets during construction is one of effective ways to avoid injuries. For the current helmet wearing detection model in a complex environment for small and dense targets, there are problems such as false detection and missed detection, an improved YOLOv5 target detection method is proposed to detect the helmet wearing. A coordinate attention mechanism(coordinate attention) is added to the backbone network of YOLOv5, which embeds location information into channel attention so that the network can pay attention on a larger area. The original feature pyramid module in the feature fusion module is replaced with a weighted bi-directional feature pyramid(BiFPN)network structure to achieve efficient bi-directional cross-scale connectivity and weighted feature fusion. The experimental results on the homemade helmet dataset show that the improved YOLOv5 model achieves an average accuracy of 95.9%, which is 5.1 percentage points higher than the YOLOv5 model, and meets the requirements for small and dense target detection in complex environments.
    Reference | Related Articles | Metrics
    Improved Sparrow Search Algorithm Based on Multi-Strategy Mixing
    HUI Lichuan, CHEN Xuelian, MENG Sibo
    Computer Engineering and Applications    2022, 58 (16): 71-83.   DOI: 10.3778/j.issn.1002-8331.2202-0134
    Abstract112)      PDF(pc) (10927KB)(321)       Save
    Dedicated to tackling the shortcomings of the simple sparrow search algorithm(SSA) with inadequate search area, sluggish convergence speed and convenient to crumple into partial top of the line when dealing with complicated optimization problems, an improved sparrow search algorithm based on multi-strategy mixing(IMSSA) is proposed. The sparrow individual position is initialized by the usage of Sine chaotic map, which enriches the vary of the population and compensates for the uneven population distribution and inadequate search space. The diversity global optimal guidance strategy with inertia weight is adopted to promote the convergence speed and regulate the overall search and local exploitation ability of the algorithm. The double-sample learning strategy is used which enables the algorithm soar out of the local optimum and enhance the population’s search capability of the solution space. The algorithm is simulated via test functions, and the effectiveness of three improved strategies is verified, as well as Wilcoxon rank sum test and time complexity evaluation have been carried out. The effects point out that the overall performance of IMSSA is notably improved. Finally, the algorithm is used to optimize the parameters of support vector machine and establish the bearing fault diagnosis model which confirms the validity of the modified strategy.
    Reference | Related Articles | Metrics
    Algorithm for Portrait Segmentation Combined with MobileNetv2 and Attention Mechanism
    WANG Xin, WANG Meili, BIAN Dangwei
    Computer Engineering and Applications    2022, 58 (7): 220-228.   DOI: 10.3778/j.issn.1002-8331.2106-0334
    Abstract162)      PDF(pc) (1064KB)(295)       Save
    As for low precision and efficiency in portrait segmentation, an algorithm for portrait segmentation combined with MobileNetv2 and attention mechanism is proposed to achieve the portrait segmentation. With keeping the encoder-decoder of U-typed network , MobileNetv2 is used as the backbone of the network and streamline the upsampling process, it can reduce the parameters of the network. It is helpful for transfer and network training. The network with attention mechanism can learn portrait features more effectively, and the mixed loss is beneficial to the classification of difficult pixels of portrait edges. A portrait bust can be selected as the input of the model, and the corresponding image mask can be produced by the network. The proposed algorithm is tested on Human_Matting dataset and EG1800 dataset. The results show that the accuracy of the proposed algorithm is 98.3%(Matting) and 97.8%(EG1800), which is higher than PortraitNet(96.3%(Matting) and 95.8%(EG1800)) and DeepLabv3+(96.8%(Matting) and 96.4%(EG1800)). The algorithm can clearly separate the target person from the background. The proposed algorithm’s IOU can reach to 98.6%(Matting) and 98.2%(EG1800), which can be used in lightweight applications and provides a new research idea for portrait segmentation.
    Reference | Related Articles | Metrics
    Computer Engineering and Applications    2022, 58 (9): 0-0.  
    Abstract80)      PDF(pc) (38025KB)(285)       Save
    Related Articles | Metrics
    Overview on Video Abnormal Behavior Detection of GAN via Human Density
    SHEN Xulin, LI Chaobo, LI Hongjun
    Computer Engineering and Applications    2022, 58 (7): 21-30.   DOI: 10.3778/j.issn.1002-8331.2110-0364
    Abstract131)      PDF(pc) (1720KB)(276)       Save
    As an important branch of computer vision, video anomaly detection is a challenging task for intelligent video surveillance systems. It is generally referred to as automatic recognition of videos that contain abnormal targets, events or behaviors, which plays a vital role in ensuring public safety. Generative adversarial network(GAN) is anemerging unsupervised method, which can not only be used to generate images, its unique adversarial learning idea also shows good development potential in the field of anomaly detection. Firstly, the framework of the GAN is introduced. Secondly, according to the density of the scene and the object on which the action is taking place, the research status of video anomaly detection based on GAN is discussed from two aspects of individual behavior anomalies, group anomalies. These two types of abnormalities are further elaborated on the basic of reconstruction and prediction methods respectively. Thirdly, the common datasets for video anomaly detection are briefly introduced, finally, the future development is prospected.
    Reference | Related Articles | Metrics
    Computer Engineering and Applications    2022, 58 (22): 0-0.  
    Abstract134)      PDF(pc) (176505KB)(273)       Save
    Related Articles | Metrics
    Research Review of Space-Frequency Domain Image Enhancement Methods
    GUO Yongkun, ZHU Yanchen, LIU Liping, HUANG Qiang
    Computer Engineering and Applications    2022, 58 (11): 23-32.   DOI: 10.3778/j.issn.1002-8331.2112-0280
    Abstract79)      PDF(pc) (1360KB)(260)       Save
    For small sample image data sets, the image enhancement method is often used to expand the amount of data to increase the rationality of the experiment. The image enhancement algorithm can improve the overall and local contrast of the image, highlight the detailed information of the image, make the image more in line with the visual characteristics of human eyes and easy to be recognized by the machine. In order to deeply study the new ideas and new directions in the application field of image enhancement, starting from the basic principle of image enhancement algorithms, based on the basic principle of image enhancement algorithms, this paper summarizes two kinds of image enhancement algorithms widely used in spatial domain and frequency domain in recent years, including histogram equalization image enhancement algorithm, gray transformation image enhancement algorithm, spatial filter image enhancement algorithm and frequency domain filter image enhancement algorithm. And their basic concepts and related definitions are introduced in detail, and their advantages and disadvantages are briefly described. In addition, subjective and objective evaluation methods are used to compare and analyze the enhancement effects of these algorithms, and the advantages and disadvantages, applicable scenarios and complexity of each algorithm are compared and analyzed, so as to further study the hidden useful information of each image enhancement algorithm, and find out the image enhancement methods with stronger robustness and applicability. Experimental results show that different algorithms have their own characteristics, for different image effects, spatial image enhancement method is more suitable for enhancing contrast, and frequency domain image enhancement method is more suitable for highlighting details. A single method can not meet the needs of image processing, and the image enhancement algorithm combined with advantages is more meaningful. The in-depth study of these algorithms can bring new opportunities for researchers, expand new research directions, promote the high-level development of the whole image enhancement technology, and make image enhancement technology play an important role in many subject fields.
    Reference | Related Articles | Metrics
    Review of Visual Odometry Methods Based on Deep Learning
    ZHI Henghui, YIN Chenyang, LI Huibin
    Computer Engineering and Applications    2022, 58 (20): 1-15.   DOI: 10.3778/j.issn.1002-8331.2203-0480
    Abstract376)      PDF(pc) (904KB)(256)       Save
    Visual odometry(VO) is a common method to deal with the positioning of mobile devices equipped with vision sensors, and has been widely used in autonomous driving, mobile robots, AR/VR and other fields. Compared with traditional model-based methods, deep learning-based methods can learn efficient and robust feature representations from data without explicit computation, thereby improving their ability to handle challenging scenes such as illumination changes and less textures. In this paper, it first briefly reviews the model-based visual odometry methods, and then focuses on six aspects of deep learning-based visual odometry methods, including supervised learning methods, unsupervised learning methods, model-learning fusion methods, common datasets, evaluation metrics, and comparison of models and deep learning methods. Finally, existing problems and future development trends of deep learning-based visual odometry are discussed.
    Reference | Related Articles | Metrics
    Review of Cognitive and Joint Anti-Interference Communication in Unmanned System
    WANG Guisheng, DONG Shufu, HUANG Guoce
    Computer Engineering and Applications    2022, 58 (8): 1-11.   DOI: 10.3778/j.issn.1002-8331.2109-0334
    Abstract260)      PDF(pc) (913KB)(244)       Save
    As the electromagnetic environment becomes more and more complex as well as the confrontation becomes more and more intense, it puts forward higher requirements for the reliability of information transmission of unmanned systems whereas the traditional cognitive communication mode is difficult to adapt to the independent and distributed development trend of broadband joint anti-interference in future. For the need of low anti-interference intercepted communications surrounded in unmanned systems, this paper analyzes the cognitive anti-interference technologies about interference detection and identification, transformation analysis and suppression in multiple domains and so on. The research status of common detection and estimation, classification and recognition are summarized. Then, typical interference types are modeled correspondingly, and transformation methods and processing problems are concluded. Furthermore, traditional interference suppression methods and new interference suppression methods are systematically summarized. Finally, the key problems of restricting the joint interference of broadband are addressed, such as the classification and recognition of unknown interference, the temporal elimination of multiple interference, the joint separation of distributed interference and the optimal control of collaborative interference, which highlight the important role of cognitive interference suppression technology in unmanned system communication.
    Reference | Related Articles | Metrics
    Multi-Scale Transformer Lidar Point Cloud 3D Object Detection
    SUN Liujie, ZHAO Jin, WANG Wenju, ZHANG Yusen
    Computer Engineering and Applications    2022, 58 (8): 136-146.   DOI: 10.3778/j.issn.1002-8331.2109-0489
    Abstract131)      PDF(pc) (1383KB)(241)       Save
    Point cloud 3D object detection has low detection accuracy for small objects such as pedestrians and bicycles, which is easy to miss detection and false detection. A 3D object detection method MSPT-RCNN(multi-scale point transformer-RCNN) based on multi-scale point cloud transformer is proposed to improve the detection accuracy of point cloud 3D objects. The method consists of two stages, the first stage(RPN) and the second stage(RCNN). In RPN stage, point cloud features are extracted through multi-scale transformer network, which includes multi-scale neighborhood embedding module and jump connection offset attention module to obtain multi-scale neighborhood geometric information and different levels of global semantic information, and generate high-quality initial 3D bounding box. In the RCNN stage, the multi-scale neighborhood geometric information of point cloud in the bounding box is introduced to optimize the position, size, orientation and confidence of the bounding box. The experimental results show that this method(MSPT-RCNN) has high detection accuracy, especially for distant and small objects. MSPT-RCNN can effectively improve the accuracy of 3D object detection by effectively learning the multi-scale geometric information in point cloud data and extracting different levels of effective semantic information.
    Reference | Related Articles | Metrics
    Research Progress in Application of Graph Anomaly Detection in Financial Anti-Fraud
    LIU Hualing, LIU Yaxin, XU Junyi, CHEN Shanghui, QIAO Liang
    Computer Engineering and Applications    2022, 58 (22): 41-53.   DOI: 10.3778/j.issn.1002-8331.2203-0233
    Abstract290)      PDF(pc) (1848KB)(240)       Save
    With the rapid development of digital finance, fraud presents new characteristics such as intellectualization, industrialization and strong concealment. And the limitations of traditional expert rules and machine learning methods are increa-
    singly apparent. Graph anomaly detection technology has a strong ability to deal with associated information, which provides new idea for financial anti-fraud. Firstly, the development and advantages of graph anomaly detection are briefly introduced. Secondly, from the perspectives of individual anti-fraud and group anti-fraud, graph anomaly detection technology is divided into individual fraud detections based on feature, proximity, graph representation learning or community division, and gang fraud detections based on dense subgraph, dense subtensor or deep network structure. The basic idea, advantages, disadvantages, research progress and typical applications of each anomaly detection technology are analyzed and compared. Finally, the common test data sets and evaluation criteria are summarized, and the development prospect and research direction of graph anomaly detection technology in financial anti-fraud are given.
    Reference | Related Articles | Metrics
    IPv6 Addressing Algorithm Using Prefix Feature
    HUANG Ping, LIU Xinlin, SUN Fengjie
    Computer Engineering and Applications    2022, 58 (11): 107-116.   DOI: 10.3778/j.issn.1002-8331.2012-0529
    Abstract35)      PDF(pc) (1230KB)(235)       Save
    With the development of the Internet and the expansion of the application of IPv6, the IP addressing engine must meet the three characteristics of high bandwidth, low search latency, and large capacity. However, the existing methods cannot simultaneously meet the above requirements. Therefore, a new IPv6 addressing algorithm is proposed here, which uses the prefix feature to construct a data structure to meet future application requirements. According to the prefix length distribution and density, it clusters them into clusters with similar characteristics, and then encodes them in a hybrid dictionary tree. The resulting data structure with memory efficiency and scalability can be stored in a low-latency memory, and allows parallelization and pipelining of the traversal process in order to support high bandwidth on the hardware. Experimental results show that the proposed algorithm reduces the amount of memory required for each prefix by 87%. In addition, when implemented on the most advanced field programmable gate array, the architecture can support processing 588 million packets per second.
    Reference | Related Articles | Metrics
    Small Object Detection Algorithm Based on Multiscale Receptive Field Fusion
    LI Chenghao, ZHANG Jing, HU Li, XIAO Xianpeng, ZHANG Hua
    Computer Engineering and Applications    2022, 58 (12): 177-182.   DOI: 10.3778/j.issn.1002-8331.2101-0009
    Abstract85)      PDF(pc) (1337KB)(227)       Save
    Aiming at the problem of low detection accuracy of general object detection algorithm in small target detection, a small object detection algorithm S-RetinaNet based on multi-scale receptive field fusion is proposed. The algorithm uses residual neural network (ResNet) to extract image features, uses recursive feature pyramid network(RFPN) to fuse features, and processes three outputs of RFPN by multiscale receptive field fusion(MRFF) to improve the ability of small target detection. Experimental results show that, compared with the original RetinaNet algorithm, the mean Average Precision(mAP) of S-RetinaNet algorithm on PASCAL VOC dataset and the average precision(AP) of MS COCO dataset are improved by 2.3 and 1.6 percentage points respectively, and the average precision small(APS) of small object detection accuracy is improved more significantly, increased by 2.7 percentage points.
    Reference | Related Articles | Metrics
    Epidermal Cell Image Recognition Research of Improved EfficientNet
    WANG Yiding, YAO Yi, LI Yaoli, CAI Shaoqing, YUAN Yuan
    Computer Engineering and Applications    2022, 58 (11): 200-208.   DOI: 10.3778/j.issn.1002-8331.2106-0186
    Abstract80)      PDF(pc) (1446KB)(225)       Save
    The amount of microscopic image data of traditional Chinese medicinal materials powder is small, and there are certain differences in features and shapes of different production areas and different collection environments. The traditional image classification methods have poor recognition results under cross-database condition. To solve the above problems, it proposes an improved method of deep convolutional neural network based on multi-channel fusion and SPP structure. First, it combines feature maps obtained by Canny edge detection and local binary pattern with original image to form a five-channel image and then it is sent to the network, in order to expand the data width of the network input; second, it embeds the improved SPP module in the EfficientNet network, in order to increase the depth of the network. The above methods can make the network pay more attention to the deep texture information of image, so that it is not affected by the collection environment, etc. , and solves the problem of cross-database recognition. The experimental results show that for two different batches of microscopic images of the epidermal cells of Chinese medicinal materials of 26 kinds, using library 1 as the training set and library 2 as the test set, the accuracy rate has increased by 2.7 percentage points to 81.5%, which proves the proposed research method has certain advantages for the task of classification of microscopic images of traditional Chinese medicinal materials under cross-database condition.
    Reference | Related Articles | Metrics
    MTICA-AEO-SVR Model for Stock Price Forecasting
    DENG Jiali, ZHAO Fengqun, WANG Xiaoxia
    Computer Engineering and Applications    2022, 58 (8): 257-263.   DOI: 10.3778/j.issn.1002-8331.2108-0433
    Abstract112)      PDF(pc) (2491KB)(215)       Save
    In order to improve the stability and separation efficiency of traditional Fast ICA algorithm, a new nonlinear function based on Tukey M estimation is constructed in this paper, and then a MTICA algorithm is obtained. Furthermore, a novel MTICA-AEO-SVR model for stock price forecasting is established combining MTICA and SVR algorithms. Firstly, the original stock data is decomposed into independent components by MTICA algorithm for sorting and denoising, and then different SVR models are selected to predict the independent components and the stock price respectively. Artificial ecosystem optimization is introduced into the SVR algorithm to select parameters, as to improve the model prediction accuracy. The empirical results of the Shanghai B-share index show that MTICA-AEO-SVR model is more accurate and efficient than ICA-AEO-SVR model and ICA-SVR model in stock price prediction.
    Reference | Related Articles | Metrics
    Summary of Intrusion Detection Models Based on Deep Learning
    ZHANG Hao, ZHANG Xiaoyu, ZHANG Zhenyou, LI Wei
    Computer Engineering and Applications    2022, 58 (6): 17-28.   DOI: 10.3778/j.issn.1002-8331.2107-0084
    Abstract244)      PDF(pc) (997KB)(215)       Save
    With the continuous in-depth development of deep learning technology, intrusion detection model based on deep learning has become a research hotspot in the field of network security. This paper summarizes the commonly used data preprocessing operations in network intrusion detection. The popular intrusion detection models based on deep learning, such as convolutional neural network, long short-term memory network, auto-encode and generative adversarial networks, are analyzed and compared. The data sets commonly used in the research of intrusion detection model based on deep learning are introduced. It points out the problems of the existing intrusion detection models based on deep learning in data set timeliness, real-time, universality, model training time and other aspects, and the possible research focus in the future.
    Reference | Related Articles | Metrics
    Research on Object Detection Algorithm Based on Improved YOLOv5
    QIU Tianheng, WANG Ling, WANG Peng, BAI Yan’e
    Computer Engineering and Applications    2022, 58 (13): 63-73.   DOI: 10.3778/j.issn.1002-8331.2202-0093
    Abstract271)      PDF(pc) (1109KB)(209)       Save
    YOLOv5 is an algorithm with good performance in single-stage target detection at present, but the accuracy of target boundary regression is not too high, so it is difficult to apply to scenarios with high requirements on the intersection ratio of prediction boxes. Based on YOLOv5 algorithm, this paper proposes a new model YOLO-G with low hardware requirements, fast model convergence and high accuracy of target box. Firstly, the feature pyramid network(FPN) is improved, and more features are integrated in the way of cross-level connection, which prevents the loss of shallow semantic information to a certain extent. At the same time, the depth of the pyramid is deepened, corresponding to the increase of detection layer, so that the laying interval of various anchor frames is more reasonable. Secondly, the attention mechanism of parallel mode is integrated into the network structure, which gives the same priority to spatial and channel attention module, then the attention information is extracted by weighted fusion, so that the network can fuse the mixed domain attention according to the attention degree of spatial and channel attention. Finally, in order to prevent the loss of real-time performance due to the increase of model complexity, the network is lightened to reduce the number of parameters and computation of the network. PASCAL VOC datasets of 2007 and 2012 are used to verify the effectiveness of the algorithm. Compared with YOLOv5s, YOLO-G reduces the number of parameters by 4.7% and the amount of computation by 47.9%, while mAP@0.5 and mAP@0.5:0.95 increases by 3.1 and 5.6 percentage points respectively.
    Reference | Related Articles | Metrics
    Overview of Smoke and Fire Detection Algorithms Based on Deep Learning
    ZHU Yuhua, SI Yiyi, LI Zhihui
    Computer Engineering and Applications    2022, 58 (23): 1-11.   DOI: 10.3778/j.issn.1002-8331.2206-0154
    Abstract199)      PDF(pc) (782KB)(196)       Save
    Among various disasters, fire is one of the main disasters that most often and universally threaten public safety and social development. With the rapid development of economic construction and the increasing size of cities, the number of major fire hazards has increased dramatically. However, the widely used smoke sensor method of fire detection is vulnerable to factors such as distance, resulting in untimely detection. The introduction of video surveillance systems has provided new ideas to solve this problem. Traditional image processing algorithms based on video are earlier proposed methods, and the recent rapid development of machine vision and image processing technologies has resulted in a series of methods using deep learning techniques to automatically detect fires in video and images, which have very important practical applications in the field of fire safety. In order to comprehensively analyze the improvements and applications related to deep learning methods for fire detection, this paper first briefly introduces the fire detection process based on deep learning, and then focuses on a detailed comparative analysis of deep methods for fire detection in three granularities:classification, detection, and segmentation, and elaborates the relevant improvements taken by each class of algorithms for existing problems. Finally, the problems of fire detection at the present stage are summarized and future research directions are proposed.
    Reference | Related Articles | Metrics
    Survey on Image Semantic Segmentation in Dilemma of Few-Shot
    WEI Ting, LI Xinlei, LIU Hui
    Computer Engineering and Applications    2023, 59 (2): 1-11.   DOI: 10.3778/j.issn.1002-8331.2205-0496
    Abstract172)      PDF(pc) (4301KB)(194)       Save
    In recent years, image semantic segmentation has developed rapidly due to the emergence of large-scale datasets. However, in practical applications, it is not easy to obtain large-scale, high-quality images, and image annotation also consumes a lot of manpower and time costs. In order to get rid of the dependence on the number of samples, few-shot semantic segmentation has gradually become a research hotspot. The current few-shot semantic segmentation methods mainly use the idea of meta-learning, which can be divided into three categories:based on the siamese neural network, based on the prototype network and based on the attention mechanism according to different model structures. Based on the current research, this paper introduces the development, advantages and disadvantages of various methods for few-shot semantic segmentation, as well as common datasets and experimental designs. On this basis, the application scenarios and future development directions are summarized.
    Reference | Related Articles | Metrics
    Field Weed Identification Method Based on Deep Connection Attention Mechanism
    SHU Yali, ZHANG Guowei, WANG Bo, XU Xiaokang
    Computer Engineering and Applications    2022, 58 (6): 271-277.   DOI: 10.3778/j.issn.1002-8331.2108-0077
    Abstract116)      PDF(pc) (1037KB)(188)       Save
    In order to achieve fast and accurate recognition of field weed images, a field weed recognition model based on deep connected attention mechanism residual network(DCECA-Resnet50-a) is proposed. Using the residual network as a benchmark, this paper improves the position of residual block downsampling, introduces the attention mechanism and connected attention mechanism modules to better extract the feature information in the images, combines the migration learning strategy to alleviate the overfitting phenomenon caused by small sample data sets, improves the generalization of the model and greatly reduces the training time of the model. The experimental results show that the improved model has the best overall performance and high recognition accuracy, with 96.31% accuracy for weeds and fewer model parameters, and achieves the accurate differentiation of four types of common weeds in pea fields, namely, silverleaf daisy, chaparral, matang and pigweed, which provides a corresponding reference for small sample data in the field of agricultural recognition.
    Reference | Related Articles | Metrics
    SSD Small Target Detection Algorithm Combining Feature Enhancement and Self-Attention
    ZHANG Xinyue, JIANG Ailian
    Computer Engineering and Applications    2022, 58 (5): 247-255.   DOI: 10.3778/j.issn.1002-8331.2109-0356
    Abstract109)      PDF(pc) (1400KB)(187)       Save
    SSD is a multi-scale target detection algorithm. Due to the lack of semantic information in shallow feature images, the detection accuracy of small targets is low. To solve this problem, a SSD small target detection algorithm, FA-SSD, which combines feature enhancement and self-attention, is proposed. The algorithm constructs a recursive reverse path from deep to shallow based on SSD, which consists of three modules:the deep feature enhancement module uses the contextual information generated from the deep multi-scale feature map and the semantic information of the deepest feature map to enhance the expression ability of the deep feature information; the up-sampling feature enhancement module enhances the semantic information of the up-sampling feature map in the reverse path by enlarging the receptive field of the feature map. The adaptive feature fusion module adaptively fuses adjacent shallow feature images and up-sampling feature images with self-attention mechanism to generate new feature images with strong semantic and precise location information. Experimental results show that on PASCAL VOC and TT100K datasets, the mAP of FA-SSD is up to 92.5% and 80.2%, indicating that this algorithm can enhance the semantic information of shallow feature images and has a good detection effect on small targets in complex scenes.
    Reference | Related Articles | Metrics
    Temporal and Spatial Fusion of Remote Sensing Images:A Review
    YANG Guangqi, LIU Hui, ZHONG Xiwu, CHEN Long, QIAN Yurong
    Computer Engineering and Applications    2022, 58 (10): 27-40.   DOI: 10.3778/j.issn.1002-8331.2111-0131
    Abstract113)      PDF(pc) (1548KB)(184)       Save
    Big data of remote sensing image with high temporal and spatial resolution plays an important role in remote sensing field. However, due to technique and budget constraints, a single satellite sensor cannot acquire remote sensing images with both high spatial resolution and high temporal resolution. Therefore, the temporal and spatial fusion technology of remote sensing image is regarded as one of the effective ways to solve the tradeoff between temporal resolution and spatial resolution. With the wide application of deep learning in various fields, deep learning technology has been proved to be a very effective method to solve image problems. According to the research results of scholars at home and abroad, the classical algorithm of remote sensing image spatiotemporal fusion is comprehensively summarized. Meanwhile, the research results of remote sensing image spatiotemporal fusion algorithm based on deep learning are analyzed, which are replicated on three datasets and the experimental results are analyzed, and the future of remote sensing image spatiotemporal fusion is prospected.
    Reference | Related Articles | Metrics
    TLS Malicious Encrypted Traffic Identification Research
    KANG Peng, YANG Wenzhong, MA Hongqiao
    Computer Engineering and Applications    2022, 58 (12): 1-11.   DOI: 10.3778/j.issn.1002-8331.2110-0029
    Abstract285)      PDF(pc) (747KB)(183)       Save
    With the advent of the 5G era and the increasing public awareness of the Internet, the public has paid more and more attention to the protection of personal privacy. Due to malicious communication in the process of data encryption, to ensure data security and safeguard social and national interests, the research work on encrypted traffic identification is particularly important. Therefore, this paper describes the TLS traffic in detail and analyzes the improved technology of early identification method, including common traffic detection technology, DPI detection technology, proxy technology, and certificate detection technology. It also introduces machine learning models for selecting different TLS encrypted traffic characteristics, as well as many recent research results of deep learning models without feature selection. The deficiencies of the related research work are summarized, and the future research work and development trend of the technology have been prospected.
    Reference | Related Articles | Metrics
    Summary of Application Research on Helmet Detection Algorithm Based on Deep Learning
    ZHANG Liyi, WU Wenhong, NIU Hengmao, SHI Bao, DUAN Kaibo, SU Chenyang
    Computer Engineering and Applications    2022, 58 (16): 1-17.   DOI: 10.3778/j.issn.1002-8331.2203-0580
    Abstract179)      PDF(pc) (967KB)(180)       Save
    Safety helmet is the most common and practical personal protective tool on the construction site, which can effectively prevent and reduce head injury caused by accidents. Helmet detection is the main work of personnel safety management on the construction site, and it is also an important content of intelligent monitoring technology on the construction site. With the development of deep learning, it has become an important part of smart site construction. In order to comprehensively analyze the research status of deep learning in helmet detection, aiming at the research of helmet detection algorithm, the commonly used helmet detection algorithm and helmet detection algorithm based on deep learning are summarized, and their advantages and disadvantages are explained in detail. On this basis, aiming at the existing problems, this paper systematically summarizes and analyzes the relevant improvement methods of helmet detection algorithm, and combs the characteristics, advantages and limitations of various methods. Finally, the future development direction of helmet detection algorithm based on deep learning is prospected.
    Reference | Related Articles | Metrics
    Real Time Traffic Sign Detection Algorithm Based on Improved YOLOV3
    WANG Hao, LEI Yinjie, CHEN Haonan
    Computer Engineering and Applications    2022, 58 (8): 243-248.   DOI: 10.3778/j.issn.1002-8331.2011-0460
    Abstract210)      PDF(pc) (625KB)(178)       Save
    Traffic sign detection is an important part of intelligent driving task. In order to meet the requirements of detection accuracy and real-time detection, an improved real-time traffic sign detection algorithm based on YOLOV3 is proposed. First, the cross stage local network is used as the feature extraction module to optimize the gradient information and reduce the inference computation. At the same time, the path aggregation network is used to replace the feature pyramid network, which not only solves the multi-scale feature fusion, but also preserves more accurate target spatial information and improves the targets detection accuracy. In addition, the complete intersection over union loss function is introduced to replace the mean square error loss to improve the positioning accuracy. Compared with other object detection algorithm on the CCTSDB dataset, experimental results show that, the average precision of the improved algorithm reaches 95.2% and the detection speed reaches 113.6 frame per second, which is 2.37% and 142% higher than YOLOV3 algorithm.
    Reference | Related Articles | Metrics
    Ranking Algorithms of Vital Nodes Based on Spring Model
    MENG Yuyu, WANG Xiao, YAN Guanghui, LUO Hao, YANG Bo, ZHANG Lei, WANG Qiong
    Computer Engineering and Applications    2022, 58 (7): 77-86.   DOI: 10.3778/j.issn.1002-8331.2103-0204
    Abstract88)      PDF(pc) (1393KB)(177)       Save
    Node ranking of vital nodes is an important problem in complex networks. When using the robustness and vulnerability of the network to evaluate the node ranking algorithms gravity model (GM) and local gravity model (LGM) based on the gravity model, once the nodes with large degrees have been removed from the network, the removal of neighbors with large gravitational values usually cannot largely affect the structure and function of the network, which shows that the algorithms still have some improvement in the ranking accuracy of vital nodes. Because of that, inspired by the spring model, further considering neighbors’ information and path information in the network, combined with the network diameter, spring model(SM) and local spring model(LSM), the node ranking algorithm and its local algorithm, are proposed. The results show that the SM algorithm and the LSM algorithm have higher accuracy for the ranking of vital nodes than other classical algorithms in synthetic networks and real networks. Especially, the SIR epidemic experiments on the Power network are conducted to furtherly verify the higher rationality and effectiveness of the SM algorithm than other algorithms.
    Reference | Related Articles | Metrics
    Convolutional Neural Network Optimization Method Based on Momentum Fractional Order Gradient Descent Algorithm
    GUO Mingxiao, WANG Hongwei, WANG Jia, LI Haozhe, YANG Shiqi
    Computer Engineering and Applications    2022, 58 (6): 80-87.   DOI: 10.3778/j.issn.1002-8331.2103-0118
    Abstract146)      PDF(pc) (1135KB)(171)       Save
    Aiming at the slow convergence speed of convolutional neural network trained by traditional gradient descent algorithm, a momentum fractional order gradient descent algorithm is proposed. The definition of fractional order calculus is introduced, and according to the problem description, the momentum thought of integer order gradient descent algorithm is applied to the fractional order gradient descent algorithm through algorithm derivation, then the momentum fractional order gradient descent algorithm is designed. The convergence of the algorithm is verified by using the test function, and the influence of different fractional orders and momentum coefficients on the algorithm are analyzed. The momentum fractional order gradient descent algorithm is used to compare with the traditional gradient descent algorithm and momentum gradient descent algorithm on three datasets. The experimental data shows that the momentum fractional order gradient descent algorithm can greatly improve the convergence speed of convolutional neural network on datasets with different complexity levels, while ensuring high classification accuracy, and saving a lot of training time. 卷积神经网络;分数阶;梯度下降;动量
    Reference | Related Articles | Metrics
    Review of Research on Small Target Detection Based on Deep Learning
    ZHANG Yan, ZHANG Minglu, LYU Xiaoling, GUO Ce, JIANG Zhihong
    Computer Engineering and Applications    2022, 58 (15): 1-17.   DOI: 10.3778/j.issn.1002-8331.2112-0176
    Abstract212)      PDF(pc) (995KB)(170)       Save
    The task of target detection is to quickly and accurately identify and locate predefined categories of objects from an image. With the development of deep learning techniques, detection algorithms have achieved good results for large and medium targets in the industry. The performance of small target detection algorithms based on deep learning still needs further improvement and optimization due to the characteristics of small targets in images such as small size, incomplete features and large gap between them and the background. Small target detection has a wide demand in many fields such as autonomous driving, medical diagnosis and UAV navigation, so the research has high application value. Based on extensive literature research, this paper firstly defines small target detection and finds the current difficulties in small target detection. It analyzes the current research status from six research directions based on these difficulties and summarizes the advantages and disadvantages of each algorithm. It makes reasonable predictions and outlooks on the future research directions in this field by combining the literature and the development status to provide a certain basic reference for subsequent research. This paper makes a reasonable prediction and outlook on the future research direction in this field, combining the literature and the development status to provide some basic reference for subsequent research.
    Reference | Related Articles | Metrics
    Research on Obstacle Avoidance Trajectory of Mobile Robot Based on Improved Artificial Potential Field
    LI Erchao, WANG Yuhua
    Computer Engineering and Applications    2022, 58 (6): 296-304.   DOI: 10.3778/j.issn.1002-8331.2108-0122
    Abstract93)      PDF(pc) (1328KB)(167)       Save
    Aiming at the problem that method of the tradition artificial potential field in the global path planning will result in inaccessibility of target, easy to fail into trap area and local minimum. A simplified obstacles and predict collision of artificial potential field method(SOPC-APF) is proposed. The concept of collision prediction is introduced that robot makes decisions before not entering trap area and local minimum. Because of the combined force of repulsion generated by multiple obstacles and the attraction of target causes the robot to fail into oscillation, simplified obstacles are proposed that means at side of the target within the influence range as restricted obstacles. Virtual target is set based on collision prediction for the problem of inaccessibility of target, and robot is guided to fast generate a smooth, stable and collision-free path by improved repulsive force function. Compared with traditional algorithm, improved APF algorithm and improved ant colony optimization algorithm, simulation experiments demonstrate that SOPC-APF can effectively solve the problem that APF is not suitable for multi-obstacle complex environment, and the traditional APF is easy to fail into trap area and local minimum.
    Reference | Related Articles | Metrics
    Historical Chinese Seal Text Recognition Based on ResNet and Transfer Learning
    CHEN Yaya, LIU Quanxiang, WANG Kaili, YI Yaohua
    Computer Engineering and Applications    2022, 58 (10): 125-131.   DOI: 10.3778/j.issn.1002-8331.2101-0247
    Abstract81)      PDF(pc) (3009KB)(165)       Save
    It is difficult to recognize historical Chinese seal text due to image degradation and super classification. In addition, the insufficient annotation data lead to poor generalization ability and classification accuracy. According to the above problems, a historical Chinese seal text recognition method based on ResNet and transfer learning is proposed. Firstly, using synthetic dataset as source domain, a pre-trained model is trained on deep residual network. Secondly, transfer learning is introduced to model combining data enhancement and label smoothing. In this process, the historical Chinese seal text dataset is taken as target domain. Finally, the recognition results in different networks are compared and the transfer learning effectiveness is analyzed. The experimental results show that this method can improve recognition accuracy effectively.
    Reference | Related Articles | Metrics
    Review of Research on Face Mask Wearing Detection
    WANG Xinran, TIAN Qichuan, ZHANG Dong
    Computer Engineering and Applications    2022, 58 (10): 13-26.   DOI: 10.3778/j.issn.1002-8331.2110-0396
    Abstract187)      PDF(pc) (733KB)(164)       Save
    Face mask wearing detection is an emerging research topic that has developed rapidly in the past two years in the context of the global COVID-19 epidemic. Under regular epidemic situation, wearing masks is an important means of effective epidemic prevention, therefore it is essential to remind and check people whether to wear masks in public places. Using artificial intelligence to complete mask wearing detection can achieve the purpose of real-time supervision, save human resources and effectively avoid mistakes, missed detection and other problems. The models and relevant algorithms used in current mask wearing detection research are reviewed. Firstly, the task and application background of mask wearing detection are described. Then, the detection algorithms based on deep neural networks and object detection models are summarized and  analyzed, the advantages and disadvantages, improvement methods and application scenarios of different research schemes are discussed. Secondly, common related data sets are introduced, and the detection performance of each algorithm is compared. Finally, the existing problems and the direction of future development are discussed and prospected.
    Reference | Related Articles | Metrics
    Lightweight Helmet Wearing Detection Algorithm of Improved YOLOv5
    YANG Yongbo, LI Dong
    Computer Engineering and Applications    2022, 58 (9): 201-207.   DOI: 10.3778/j.issn.1002-8331.2111-0346
    Abstract197)      PDF(pc) (1476KB)(164)       Save
    Aiming at the problems of the existing helmet wearing detection algorithm, such as multiple parameters, complex network, large amount of calculation, which is not suitable for deployment on embedded devices, and poor discrimination of occlusion targets, an improved lightweight helmet detection algorithm, YOLo-M3, is proposed.?Firstly, the YOLOv5s backbone network is replaced by MobileNetV3 for feature extraction, which reduces the number of parameters and computation of the network.?Secondly, Diou-NMS is used to replace NMS to improve the identification of occlusion targets. CBAM attention mechanism is added to make the model pay more attention to the main information to improve the detection accuracy. Finally, knowledge distillation is carried out to increase the recall rate and accuracy of model detection.?Experiments verify that YOLO-M3 algorithm can improve the identification of occlusion targets, and reduce the calculation amount of YOLOv5s model by 42% and the model size by 40% while ensuring a high average detection accuracy, thus reducing the hardware cost and meeting the requirements of deployment in embedded end.
    Reference | Related Articles | Metrics
    Random Forest Algorithm Based on PCA and Hierarchical Selection Under Spark
    LEI Chen, MAO Yimin
    Computer Engineering and Applications    2022, 58 (6): 118-127.   DOI: 10.3778/j.issn.1002-8331.2009-0316
    Abstract78)      PDF(pc) (1476KB)(164)       Save
    In the context of big data, the random forest algorithm has large covariance matrix, insufficient coverage of subspace feature information and high node communication overhead. A parallel random forest algorithm based on PCA and subspace hierarchical selection, PLA-PRF(PCA and subspace layer sampling on parallel random forest algorithm). For the initial feature set, a PCA-based matrix factorization strategy(MFS) is proposed to extract principal component features to solve the problem of large covariance matrix in the process of feature transformation. Based on the obtained principal component features, a hierarchical subspace construction algorithm(error-constrained hierarchical subspace construction algorithm, EHSCA) based on error constraints is proposed, which selects pheromone features hierarchically, constructs feature subspaces, and solves the problem of insufficient coverage of subspace feature information. In the process of parallel training decision trees in the Spark environment, a data reuse strategy(DRS) is designed to solve the problem of high node communication overhead. By vertically dividing RDD data objects, it improves the performance of the distributed environment. Data utilization rate solves the problem of high node communication overhead. Experimental results show that PLA-PRF has better classification effect and higher parallelization efficiency.
    Reference | Related Articles | Metrics
    Survey of Opponent Modeling Methods and Applications in Intelligent Game Confrontation
    WEI Tingting, YUAN Weilin, LUO Junren, ZHANG Wanpeng
    Computer Engineering and Applications    2022, 58 (9): 19-29.   DOI: 10.3778/j.issn.1002-8331.2202-0297
    Abstract529)      PDF(pc) (904KB)(164)       Save
    Intelligent game confrontation has always been the focus of artificial intelligence research. In the game confrontation environment, the actions, goals, strategies, and other related attributes of agent can be inferred by opponent modeling, which provides key information for game strategy formulation. The application of opponent modeling method in competitive games and combat simulation is promising, and the formulation of game strategy must be premised on the action strategy of all parties in the game, so it is especially important to establish an accurate model of opponent behavior to predict its intention. From three dimensions of connotation, method, and application, the necessity of opponent modeling is expounded and the existing modeling methods are classified. The prediction method based on reinforcement learning, reasoning method based on theory of mind, and optimization method based on Bayesian are summarized. Taking the sequential game(Texas Hold’em), real-time strategy game(StarCraft), and meta-game as typical application scenarios, the role of opponent modeling in intelligent game confrontation is analyzed. Finally, the development of adversary modeling technology prospects from three aspects of bounded rationality, deception strategy and interpretability.
    Reference | Related Articles | Metrics
    Review of Few-Shot Object Detection
    ZHANG Zhenwei, HAO Jianguo, HUANG Jian, PAN Chongyu
    Computer Engineering and Applications    2022, 58 (5): 1-11.   DOI: 10.3778/j.issn.1002-8331.2109-0405
    Abstract167)      PDF(pc) (1012KB)(163)       Save
    Recently, object detection based on deep learning has been achieved remarkable achievements and various of mature models have been proposed. However, most of these models rely on a large number of annotated training samples. Besides, in practical applications, it is often difficult to get access to large scale of high-quality annotated samples, which limits its application and popularization in specific areas. Few-shot object detection has been extensively researched taking advantage of its small dependence on the number of samples. Based on the current research, this paper reviews the current mainstream of the few-shot object detection systematically, including problem definition, mainstream methods, as well as common experimental designs. Then, it points out potential application directions. Furthermore, the generalized few-shot object detection is also briefly introduced. Finally, the paper analyzes challenges of the few-shot object detection technology and discusses corresponding countermeasures.
    Reference | Related Articles | Metrics
    Complex Road Target Detection Algorithm Based on Improved YOLOv5
    WANG Pengfei, HUANG Hanming, WANG Mengqi
    Computer Engineering and Applications    2022, 58 (17): 81-92.   DOI: 10.3778/j.issn.1002-8331.2205-0158
    Abstract163)      PDF(pc) (1217KB)(161)       Save
    Aiming at the problem of false detection and missed detection caused by dense occluded targets and small targets in complex road background, a complex road target detection algorithm based on improved YOLOv5 is proposed. Firstly, Quality Focal Loss is introduced, which combines the classification score with the quality prediction of location to improve the positioning accuracy of dense occluded targets. Secondly, a shallow detection layer is added as the detection layer of smaller targets, the three-scale detection of the original algorithm is changed to four-scale detection, and the feature fusion part is also improved accordingly, which improves the learning ability of the algorithm to the features of small targets. Then, based on the feature fusion idea of weighted bidirectional feature pyramid network(BiFPN), a de-weighted BiFPN is proposed, which makes full use of deep, shallow and original feature information, strengthens feature fusion, reduces the loss of feature information in the process of convolution, and improves the detection accuracy. Finally, the convolution block attention module(CBAM) is introduced to further improve the feature extraction ability of the algorithm and make the algorithm pay more attention to useful information. The experimental results show that the detection accuracy of the improved algorithm in this paper on the public autopilot data set KITTI and the self-made rider helmet data set Helmet reaches 94.9% and 96.8% respectively, which is 1.9 percentage points and 2.1 percentage points higher than the original algorithm, and the detection speed reaches 69 FPS and 68 FPS respectively. It has better detection accuracy and real-time performance. At the same time, compared with some mainstream target detection algorithms, the improved algorithm in this paper also has some advantages.
    Reference | Related Articles | Metrics
    Survey of IoT Forensics
    LIANG Guangjun, XIN Jianfang, WANG Qun, NI Xueli, GUO Xiangmin, XIA Lingling
    Computer Engineering and Applications    2022, 58 (8): 12-32.   DOI: 10.3778/j.issn.1002-8331.2201-0014
    Abstract169)      PDF(pc) (1143KB)(159)       Save
    The advent of the Internet of Things era brings great convenience to people, but it also makes the scope of cyberspace attacks wider and brings new cyberspace security threats. Massive IoT devices retain a wealth of digital traces, which can provide insight into people’s various behaviors at home and other places. This is of great significance for digital forensics. This article has an in-depth discussion on IoT forensics, starting with the rise, development and research status of IoT forensics, and further discusses digital forensics models, 1-2-3 regional method models, parallel structure-IoT forensics models, privacy protection forensics models, and forensic model for special applications. Finally, it elaborates on the opportunities and challenges of forensics in the Internet of Things. This article strives to provide readers with help and reference for more in-depth research on the basis of systematically learning IoT forensics technology.
    Reference | Related Articles | Metrics
    Image Semantic Segmentation Based on Fully Convolutional Neural Network
    ZHANG Xin, YAO Qing’an, ZHAO Jian, JIN Zhenjun, FENG Yuncong
    Computer Engineering and Applications    2022, 58 (8): 45-57.   DOI: 10.3778/j.issn.1002-8331.2109-0091
    Abstract175)      PDF(pc) (1057KB)(159)       Save
    Image semantic segmentation is a hot research topic in the field of computer vision. With the rapid rise of fully convolutional neural networks, the development of fusion of image semantic segmentation and fully convolutional networks has shown very bright results. Through the collection of high-quality literature in recent years, the focus is on the summary of full convolutional neural network image semantic segmentation methods. The collected literature is divided into classical semantic segmentation, real-time semantic segmentation and RGBD semantic segmentation according to the application scenarios, and then the representative segmentation methods are described. Commonly used public datasets and evaluation metrics for performance are also summarized, and experiments on commonly used datasets are analyzed and summarized. Finally, the possible future research directions of fully convolutional neural networks are prospected.
    Reference | Related Articles | Metrics
    CA-YOLOv5 for Crowded Pedestrian Detection
    CHEN Yixiao, Alifu·Kuerban, LIN Wenlong, YUAN Xu
    Computer Engineering and Applications    2022, 58 (9): 238-245.   DOI: 10.3778/j.issn.1002-8331.2201-0058
    Abstract200)      PDF(pc) (1115KB)(155)       Save
    Aiming at the problem of high miss-detection rate and insufficient feature fusion of YOLOv5 in crowded pedestrian detection task, the CA-YOLOv5 pedestrian detection algorithm is proposed. To solve the problem of insufficient fine-grained feature fusion in the original backbone network, Res2Block is used to rebuild the backbone network of YOLOv5, so as to improve the fine-grained feature fusion ability of the network and improve the detection accuracy. For the large change of target scale in dataset, coordinate attention is introduced to enhance the receptive field and the model’s ability to accurately locate the target. Aiming at the problem that FPN structure reduces the multi-scale feature expression ability during feature fusion, the feature enhancement module is proposed to enhance the multi-scale feature expression ability. Through the structural re-parameterization method to reduce the number of parameters and computation in the model, and speed up the detection. Aiming at the common problem of crowded pedestrians in pedestrian detection task, EViT is proposed to enhance the ability of the model to pay attention to local information and improve the detection accuracy. Experimental results show that in the crowded pedestrian detection task, the detection accuracy of CA-YOLOv5 reaches 84.86%, 3.75% higher than the original algorithm, and the detection speed can reach 51?FPS, which has good detection accuracy and real-time. Therefore, it can be better applied to real-time pedestrian detection task.
    Reference | Related Articles | Metrics