Content of Engineering and Applications in our journal

Published in last 1 year |  In last 2 years |  In last 3 years |  All Please wait a minute... For Selected: Download Citations EndNote Ris BibTeX Toggle Thumbnails Select Risk Propagation and Intervention in Complex Supply Chain Networks During Unforeseen Public Events JIANG Lin, LIANG Jingxin Computer Engineering and Applications    2024, 60 (8): 296-308.   DOI: 10.3778/j.issn.1002-8331.2306-0327 Abstract （9）      PDF（pc） （1095KB）（4）       Knowledge map    Save Unforeseen public events present substantial challenges to the stable operation of supply chain networks. It is of paramount importance to enhance the capacity of these networks to respond to unforeseen public events while preserving their resilience and stability. In consideration of the unique characteristics of such events, this paper formulates a risk propagation model for intricate supply chain networks under the influence of government intervention, utilizing an enhanced SEIR system dynamics transmission model. It employs the basic reproduction number theory to scrutinize the thresholds and equilibrium points in risk propagation, delves into the constructive role of government intervention in disrupting the proliferation of supply chain risks, and furnishes a macro-level analysis of the mechanisms governing risk propagation within supply chains and the accompanying intervention strategies, ultimately offering directives for government risk management. Both theoretical research and simulation experiments concur in affirming that, through the manipulation of the basic reproduction number, the government can effectively curtail the further expansion and diffusion of risks, thereby attaining the objective of risk containment. Reference | Related Articles | Metrics Select Research on Evacuation Path Planning in Fire Environment with Improved Ant Colony Algorithm DU Yun, LIU Xiaoyu, JIA Kejin, DING Li, HUANG Gongfa Computer Engineering and Applications    2024, 60 (8): 309-319.   DOI: 10.3778/j.issn.1002-8331.2310-0416 Abstract （10）      PDF（pc） （1036KB）（6）       Knowledge map    Save Aiming at the problem of building fire personnel evacuation, a path planning model with improved ant-colony algorithm is proposed to ensure the safety of fire personnel evacuation.When the search direction is close to the target node, the increase of the value of the directional information function makes the pheromone different. According to the fire impact degree, the fire grade function is established, so that the transfer probability decreases with the increase of the fire grade, and the blindness of the ant colony in finding the way is reduced. By analyzing the influence factors of fire, the equivalent length is established and the heuristic function is constructed to avoid falling into the local optimal. The volatili-zation coefficient of pheromone is adjusted adaptively with the fire grade function to accelerate the volatilization rate of the fire path pheromone and improve the global search ability of the algorithm. At the same time, the reward and punishment coefficient and fuzzy control are introduced in the pheromone updating strategy to improve the robustness and path smoothness of the evacuation system, and the global pheromone is restricted to balance the local development and global search ability of the algorithm. The simulation results show that the improved ant colony algorithm can efficiently plan evacuation routes in the case of fire or not. Reference | Related Articles | Metrics Select Safety Helmet Wearing Detection Algorithm for Distribution Network Construction in Natural Scenarios XU Kui, LI Xinzhuo, ZHANG Li, ZHANG Junjie, YANG Ning Computer Engineering and Applications    2024, 60 (8): 320-328.   DOI: 10.3778/j.issn.1002-8331.2301-0057 Abstract （13）      PDF（pc） （690KB）（11）       Knowledge map    Save For high-risk industries such as distribution network construction operations, wearing safety helmets in accordance with safety codes during construction is one of the effective ways to avoid accidents. Due to the complex and changeable construction environment of distribution network, the existing safety helmet identification methods often have the problem of false detection and leakage in natural scenarios and cannot meet the real-time requirements. In order to improve the recognition accuracy and efficiency of safety helmets in natural scenes, a safety helmet wearing recognition detection network model YOLO-ACON-Attention for distribution network construction in natural scenes is proposed. Based on the YOLOv5 algorithm, the adaptive judgment activation function is used to replace the original activation function to strengthen the model detection ability. Secondly, the adaptive attention module is constructed by using the two-round and four-way IRNN network in the backbone network to improve the image information feature extraction ability of the model. Experimental results show that compared with the original YOLOv5 algorithm, the accuracy and recall of the algorithm are 94.75% and 89.29%, which are improved by 7.65% and 5.17%, respectively, and the detection speed is 36.5 FPS. Reference | Related Articles | Metrics Select Research on Autonomous Positioning and 3D Map Building of Indoor Mapping Robots ZHOU Hongyi, ZHANG Guobao, ZHU Hongwei Computer Engineering and Applications    2024, 60 (8): 329-337.   DOI: 10.3778/j.issn.1002-8331.2301-0073 Abstract （8）      PDF（pc） （873KB）（3）       Knowledge map    Save Indoor space is difficult to build 3D maps with high accuracy using limited sensors due to low illumination, lack of GPS positioning assistance and few scene features. To address this problem, this paper improves the FAST-LIO algorithm by introducing a Lider-IMU parameter initialization system and a back-end loopback detection optimization algorithm to increase the robustness of map building in large scenes. Experimental studies are conducted using publicly available datasets. The results show that the trajectory error accuracy of the algorithms in this paper are improved compared with the existing algorithms. This paper also designs a robot to test in the interior environment of a building at Southeast University. The experimental results show that the robot can realize the autonomous movement to build the map and return safely with good results. Reference | Related Articles | Metrics Select Hardware Accelerator Supporting Inhibitory Spiking Neural Network QIAN Ping, HAN Rui, XIE Lingdong, LUO Wang, XU Huarong, LI Songsong, ZHENG Zhendong Computer Engineering and Applications    2024, 60 (8): 338-347.   DOI: 10.3778/j.issn.1002-8331.2301-0133 Abstract （9）      PDF（pc） （609KB）（6）       Knowledge map    Save The design of existing spiking neural network accelerators pays too much attention to the functional integrity of the hardware level and lacks relevant collaborative optimization at the algorithm level to ensure hardware computing efficiency. In addition, traditional event-driven spiking neural network accelerators do not consider the ubiquitous spike jitter phenomenon in spiking neuron models, so they cannot support inhibitory spiking neural networks. In order to solve the above problems, a design method of a suppressive spiking neural network accelerator is proposed by combining software and hardware. At the software optimization level, through the analysis of the calculation redundancy of the spiking neural network, a corresponding approximate calculation method is proposed to reduce the calculation amount of the spiking neural network greatly; at the hardware design level, a calculation module to solve the problem of pulse jitter is proposed, and on this basis, a parallel computing structure suitable for the approximate computing method is designed. In order to verify the rationality of the design, the accelerator prototype FEAS is deployed on Xilinx ZC706 FPGA. The test results on mainstream datasets show that compared with the previous accelerator deployment of spiking neural networks, FEAS has achieved more than an order of magnitude performance improvement while maintaining 97.54% of the original model accuracy. Reference | Related Articles | Metrics Select Immediate Prediction Model of SPAD Value and Maturity of Fresh Tobacco Leaves in Field PEI Wencan, SUN Guangwei, HUANG Jinguo, XU Dinghui, LIU Jing Computer Engineering and Applications    2024, 60 (8): 348-360.   DOI: 10.3778/j.issn.1002-8331.2301-0153 Abstract （7）      PDF（pc） （840KB）（6）       Knowledge map    Save As a key link in the process of tobacco harvest, maturity measurement should be immediate, scientific and accurate. There are some problems in the current implementation of tobacco maturity recognition. For example, the measurement instruments are expensive and cumbersome to operate, which cannot be promoted for use in the field; the embedded image processing algorithms in phone cameras interfere with the effective features of images and the complex weather environment in the field affects the consistency of image acquisition; existing recognition algorithms ignore botanical domain information, which affects model accuracy and universality. Accordingly, this paper proposes a low-cost and effective method to predict the SPAD value and the maturity of fresh tobacco leaves in the field immediately, which ensures the quality of subsequent modulation of tobacco leaves by improving the prediction accuracy. Firstly, this paper develops a portable shooting device to achieve high-quality image acquisition in the field and proposes a segmentation method adapted to tobacco images in the field taking the CX-26 as the research object. After that, this paper extracts the feature data of the image target area and builds a SPAD value prediction model and maturity identification model in turn using the XGBoost algorithm. Then, a model integration idea for two models is proposed, which is able to use the strong correlation between SPAD value and maturity to improve maturity identification accuracy by predicting SPAD values. The method performs well in all evaluation indexes, where the mean absolute error reaches 0.470 3 for SPAD value prediction and macro F1-Score reaches 95.27% for maturity identification. Finally, this paper develops an APP to realize the transmission of tobacco images and prediction results between the portable shooting device and the model to achieve fast, objective and accurate immediate prediction of tobacco maturity in the field. The results can provide effective technical support for the accurate harvest of crops in the field. Reference | Related Articles | Metrics Select Stock Price Prediction Integrating Investor Sentiment Based on S_AM_BiLSTM Model YUAN Jing, PAN Su, XIE Hao, XU Wenpeng Computer Engineering and Applications    2024, 60 (7): 274-281.   DOI: 10.3778/j.issn.1002-8331.2305-0059 Abstract （33）      PDF（pc） （729KB）（31）       Knowledge map    Save Stock price forecasting has always been one of the research hotspots in the financial field. However, the formation mechanism of stock price is quite complex, and various factors may lead to the change of stock price. Therefore, this paper proposes a hybrid model of stock price forecasting based on deep learning method and integrating multi-source data and investor sentiment (S_AM_BiLSTM). First, it uses the TextCNN to analyze the sentiment of investor comments extracted from stock forums, and calculates the emotional factors. Then, taking emotional factors, technical indicators and stock historical trading data as the feature set of stock price prediction, the bi-directional long short-term memory neural network is used to predict the stock closing price, and on this basis, attention mechanism is added to improve the prediction accuracy. In order to prove the validity and applicability of the model, it randomly selects four key industries for empirical research. The experimental results show that the proposed hybrid model is more effective than other single models and models without emotional factors. Reference | Related Articles | Metrics Select Research on Steel Surface Defect Detection with Improved YOLOv7 Algorithm GAO Chunyan, QIN Shen, LI Manhong, LYV Xiaoling Computer Engineering and Applications    2024, 60 (7): 282-291.   DOI: 10.3778/j.issn.1002-8331.2308-0414 Abstract （59）      PDF（pc） （1101KB）（63）       Knowledge map    Save At present, the intelligent inspection technology based on deep learning is gradually applied to the field of steel surface defect detection. Aiming at the problem of low precision of steel surface defect detection, a high-precision and real-time defect detection algorithm CDN-YOLOv7 is proposed. Firstly, CARAFE lightweight up-sampling operator is added to improve the feature fusion capability of the network. Then, the YOLOv7 detection head network is redesigned by integrating the cascade attention mechanism and decoupling heads, aiming to solve the problem of low feature utilization efficiency of the original head network and make full use of multi-dimensional information of different scales, channels and spaces, improve the ability of model representation in complex scenarios. Finally, normalized Wasserstein distance is introduced to redesign Focal-EIoU loss function, and NF-EIoU is proposed to replace CIoU loss, balance the contribution of defect samples at different scales to loss, and reduce the missed detection rate of defects at different scales. The experimental results show that the detection accuracy of CDN-YOLOv7 can reach 80.3%, which is 6.0 percentage points higher than that of the original YOLOv7, and the model reasoning speed can reach 60.8 frame/s, meeting the real-time requirements. While improving the detection accuracy of defects at various scales, CDN-YOLOv7 significantly reduces the missed detection rate of defects. Reference | Related Articles | Metrics Select Research on Policy Tools Classification Based on ChatGPT Augmentation and Supervised Contrastive Learning HU Zhiqiang, LI Pengjun, WANG Jinlong, XIONG Xiaoyun Computer Engineering and Applications    2024, 60 (7): 292-305.   DOI: 10.3778/j.issn.1002-8331.2308-0354 Abstract （26）      PDF（pc） （876KB）（30）       Knowledge map    Save The classification of policy tools is an important dimension in the quantification and analysis of policy texts. Due to the scarcity of training data, models are prone to overfitting, resulting in reduced prediction confidence and an increased risk of misclassification. Therefore, policy tool classification method based on ChatGPT augmentation and supervised contrastive learning is proposed. The method consists of two stages：pre-training language model fine-tuning and ChatGPT decision augmentation. In the first stage, ChatGPT, a large language model, augments policy texts to increase the training dataset, while supervised contrastive learning fine-tune the RoBERTa model to improve classification performance. In the second stage, ChatGPT assists in the decision-making process for low confidence texts through the pre-trained language model and reduces the risk of misclassifying similar texts. Experiments on the digital industry policy tools classification dataset and the Tnews dataset show that the proposed method surpasses mainstream research approaches and can effectively improve the performance of the base model, with a more significant improvement observed when the training samples are limited. Reference | Related Articles | Metrics Select Research on Optimization of YOLOv5s Detection Algorithm for Steel Surface Defect XU Hongjun, TANG Ziqiang, ZHANG Jindong, ZHU Peihua Computer Engineering and Applications    2024, 60 (7): 306-314.   DOI: 10.3778/j.issn.1002-8331.2307-0275 Abstract （66）      PDF（pc） （805KB）（59）       Knowledge map    Save Aiming at the problems that YOLOv5 has insufficient ability to extract complex features of steel defects and the detect results are susceptible to background environment, a steel surface defect detection algorithm based on YOLOv5s is proposed. This algorithm introduces CBAM attention into C3 to enhance attention to key information. It utilizes the CARAFE to replace the nearest neighbor interpolation algorithm, reducing the loss of feature information caused by upsampling. It proposes to replace the SPPF in YOLOv5 with the SPPCPSC, which can improve the expressive ability of the network. The experimental results show that the mAP@0.5 of the proposed YOLOv5s improved model on the NEU-DET dataset reaches 76.6%, which is 2.3 percentage points higher than that of YOLOv5s. The model parameters are basically the same as YOLOv5s. The CARAFE module is the main reason for the slowdown of the improved model detection speed. In addition, the combination of the CARAFE and the SPPCSPC_group has a good effect on the detection accuracy of the model. Reference | Related Articles | Metrics Select Drug Recommendation Model for Graph Embedding Dual Graph Convolutional Network JIANG Yuzhe, CHENG Quan Computer Engineering and Applications    2024, 60 (7): 315-324.   DOI: 10.3778/j.issn.1002-8331.2211-0402 Abstract （22）      PDF（pc） （646KB）（19）       Knowledge map    Save In recent years, drug recommendation based on deep learning models has been extensively studied and widely applied in the field of wisdom medicine. This paper proposes a drug recommendation model with dual graph convolutional network based on graph embedding. Firstly, it constructs the knowledge graph of patient’s attributes and medications. The embedding representation is obtained using graph embedding. Secondly, the embedding representation of the knowledge graph of patient’s attributes are put into the multilevel graph attention network layer loaded with attention mechanism and bidirectional propagation mechanism for disseminating and aggregating information. Then, the representation of patient’s attributes and the embedded representation of the knowledge graph of patient’s medications are integrated. They are put into the multilevel graph attention network layer training again to mine the high-level association between patient’s attributes and medications. Finally, the drug recommendation is completed. It carries out an empirical study with the basic patient’s information, physiological characteristics and patient’s medication data in the data set of the medical information mart for intensive care Ⅳ as the objects. The experimental results prove that it outperforms the baseline method in four evaluation indexes：precision, recall, F1 score and NDCG. Reference | Related Articles | Metrics Select Nomad Algorithm with Constraints Research on Bike-Sharing Allocation GUO Maozu, MA Li, ZHAO Lingling Computer Engineering and Applications    2024, 60 (7): 325-334.   DOI: 10.3778/j.issn.1002-8331.2212-0047 Abstract （26）      PDF（pc） （626KB）（26）       Knowledge map    Save The bike-sharing allocation is an important way to optimize the urban traffic resources rebalancing, but the current optimal-route allocation method is sensitive to the bike system magnitude. Therefore, a time-based and inter-regional bike-sharing allocation method is researched, and the nomad algorithm with constraints (NCA) is proposed to obtain the optimal allocation solution. Firstly, with the bike flow as the constraints and the minimal operation loss as the target, the allocation problem is modeled as a multi-constrained objective optimization problem. Then, NCA is proposed to predict the optimal bike inventory in the stations and the transfer amount among the stations. Compared with the original nomadic algorithm without constraint thinking, NCA improves the local search strategies and the global optimization strategies, and optimizes the tribe generation methodology. Finally, based on the predicted inventory and transfer amount, the interregional allocation scheme in different time periods is obtained. The comparative experimental results on the relevant datasets in Shanghai and New York show that the running time is about 15% of other methods. The demand response rate is 0.15% higher than the branch-and-bound algorithm. The bike quantity and the operating losses are reduced by about 10% compared to the genetic algorithm. It can be seen that the proposed method has higher optimization efficiency and user demand response rate. Reference | Related Articles | Metrics Select Research on Urban Logistics Distribution Mode of Bus-Assisted Drones PENG Yong, REN Zhi Computer Engineering and Applications    2024, 60 (7): 335-343.   DOI: 10.3778/j.issn.1002-8331.2212-0252 Abstract （29）      PDF（pc） （755KB）（32）       Knowledge map    Save The rapid development of e-commerce forces the continuous transformation and upgrading of the logistics industry. In view of the fact that local governments encourage the development of public transport and advocate green and low-carbon logistics distribution mode, a distribution mode of bus-assisted drone is studied. After explaining the problem, a mathematical model with the lowest distribution cost is constructed, and a heuristic algorithm of smart general variable neighborhood search metaheuristic is designed to solve the problem. At the same time, in order to improve the efficiency of the algorithm, K-means clustering and greedy algorithm are introduced to generate the initial solution. Firstly, aiming at different scale examples, a variety of local search strategies and a variety of algorithms are compared to verify the effectiveness of the algorithm. Secondly, by selecting the standard CVRP as example, the single truck distribution mode and truck-drone collaborative distribution mode are compared with the distribution mode of bus-assisted drone to prove its cost and time advantages. Finally, Beijing Bus Rapid Transit Line 2 and its surrounding customer points are selected, and sensitivity analysis is made by changing the bus stop spacing and departure interval, result shows that the impact of increasing the stop spacing is greater than the change of departure interval. Reference | Related Articles | Metrics Select Ship Pipe Layout Design Based on Improved Ant Colony Optimization DONG Zongran, CHEN Heng, BIAN Xuanyi, LOU Oujun Computer Engineering and Applications    2024, 60 (7): 344-354.   DOI: 10.3778/j.issn.1002-8331.2212-0298 Abstract （42）      PDF（pc） （729KB）（63）       Knowledge map    Save In order to provide engineers with a variety of high-quality ship pipe layouts in a relatively short time under multiple constraints and objectives, a novel ship pipe layout design method based on the modified ant colony optimization (MdACO) is proposed. The method works in a grid-decomposition space model, firstly the optimization objectives of the fitness function are normalized, and then the key steps of MdACO are improved, such as the selection of ant-moving direction, the mechanism of pheromone update, and the procedure of ant route searching. The best solution set is used to save lots of optimization individuals with the same fitness but different layout effects. Moreover, the auxiliary connection point and parallel computing are integrated to improve the searching ability and efficiency of the algorithm. Finally, the effectiveness and advancement of the algorithm are verified by the simulation test. Reference | Related Articles | Metrics Select Research on Fine-Grained Fault Diagnosis of Rolling Bearings RUAN Hui, HUANG Xixia, LI Dengfeng, WANG Le Computer Engineering and Applications    2024, 60 (6): 312-322.   DOI: 10.3778/j.issn.1002-8331.2210-0341 Abstract （29）      PDF（pc） （895KB）（33）       Knowledge map    Save Aiming at the current situation that supervised deep learning is mainly used to extract fault features and detect coarse-grained types of faults in rolling bearing fault diagnosis, a fine-grained fault diagnosis method for rolling bearings integrated with Gaussian mixture models (GMM) and deep residual shrinkage networks (DRSN) is proposed. The GMM model integrates multiple Gaussian distribution functions to fit the distribution of fine-grained fault data and realize the clustering of bearing vibration signals without labels. The attention mechanism in DRSN model focused on the more critical information for the current task from a large number of fault feature information. Soft threshold is designed to set different thresholds for bearing samples in different health states. The method is validated by collecting 30 bearing health states from Case Western Reserve University (CWRU) data. The results show that integrate the unsupervised model with the deep learning model, which can process the bearing fault data without labels, achieve the purpose of fine-grained classification of bearing faults, provide a basis for subsequent equipment maintenance, and have good practical engineering significance and popularization. Reference | Related Articles | Metrics Select Human Respiratory Rate Detection on UAV Platform LIANG Shuai, YANG Xuezhi, ZANG Zongdi Computer Engineering and Applications    2024, 60 (6): 323-329.   DOI: 10.3778/j.issn.1002-8331.2211-0159 Abstract （24）      PDF（pc） （698KB）（28）       Knowledge map    Save It is a means of injury assessment to detect respiratory rate using an unmanned aerial camera. However, the existing video-based respiratory rate detection algorithms are only applicable to fixed cameras. On the basis of spatial phase based respiratory signal extraction technology, a non-contact measurement method of human respiratory rate with videos recorded by unmanned aerial vehicle (UAV) is proposed. The complex steerable pyramid is used to extract the spatial phase of each frame image, and the phase sequence is obtained in chronological order. Secondly, the empirical mode decomposition (EMD) is used to decompose multiple modal components from the phase sequence, and the frequency variability analysis model is designed to select the components with stable frequencies, or the target respiratory signal. The peak value detection method is used to detect the human respiratory rate. The experimental results show that the average accuracy of the method can reach more than 98%, which is superior to the existing detection methods. Reference | Related Articles | Metrics Select Hybrid LightGBM Model for Breast Cancer Diagnosis XING Changzheng, XU Jiayu Computer Engineering and Applications    2024, 60 (6): 330-338.   DOI: 10.3778/j.issn.1002-8331.2211-0218 Abstract （34）      PDF（pc） （692KB）（53）       Knowledge map    Save Breast cancer is one of the most common types of cancer, and its prevalence continues to rise every year. Without surgical biopsy, it can effectively provide auxiliary diagnosis and treatment for doctors and reduce the pain of patients by analyzing various indicators of the nucleus to predict whether the mass is benign or not. Therefore, a breast cancer diagnosis model based on LightGBM algorithm is proposed. Firstly, the borderline-synthetic minority oversampling technique (Borderline-SMOTE)  is used to improve the problem of imbalanced breast cancer diagnosis data. Secondly, the PWLCM chaotic map, the new inertia weight and the criss-cross algorithm are introduced into the sparrow search algorithm (SSA)  to improve it, and then the improved SSA algorithm is used to automatically optimize the parameters of LightGBM. Then, because LightGBM is sensitive to noise, an OVR-Jacobian regularization method is proposed to reduce the noise of LightGBM. Finally, the improved LightGBM hybrid model is used to diagnose breast cancer. The experimental results show that the proposed hybrid model is superior to the common models in the three indicators of mean square error, coefficient of determination and cross-validation score, showing its better diagnostic effect. Reference | Related Articles | Metrics Select Path Planning for Blood Transportation by Unmanned Aerial Vehicles with Safety of Taking-Off and Landing XU Weihua, ZHANG Genrui, QIU Longlong, ZHAO Caimei, XIONG Jian Computer Engineering and Applications    2024, 60 (6): 339-348.   DOI: 10.3778/j.issn.1002-8331.2211-0296 Abstract （50）      PDF（pc） （692KB）（52）       Knowledge map    Save To solve the problem of path planning in blood transportation by unmanned aerial vehicles (UAVs), a path planning model for blood transportation by UAVs is established with the goal of minimizing total transportation distance of UAVs. Considering the safe time interval of continuous take-off and landing of UAVs under the condition that the number of UAVs take-off and landing platforms is limited, a scheduling strategy of UAVs take-off sequence is designed to reduce the total time consumption for UAVs to complete transportation. And an imperialist competitive algorithm based on the imperialist reform is proposed. A sinusoidal disturbance strategy and an imperialist reform are introduced to improve the search accuracy of the algorithm. The acceptance criteria related to solution quality is designed to increase the diversity of the population. The benchmark example and the UAVs transporting blood example are used for verification. The results show that the proposed algorithm can provide transport tasks for the blood transportation by UAVs that meets various constraints and without UAVs take-off and landing conflicts. In addition, using the scheduling strategy of UAVs take-off sequence can reduce the total time spent by UAVs to actually complete the tasks effectively. Reference | Related Articles | Metrics Select Path Tracking Control Strategy for 4WD/4WS Autonomous Vehicle with Considering Lateral Stability XIANG Jingyan, ZHOU Kui, FU Yongzhi, XU Yang, PENG Xufeng Computer Engineering and Applications    2024, 60 (6): 349-358.   DOI: 10.3778/j.issn.1002-8331.2211-0154 Abstract （36）      PDF（pc） （884KB）（57）       Knowledge map    Save According to the characteristics of single-wheel solo controllable of unmanned vehicle with distributed four-wheel drive/steering, a path tracking control strategy considering stability is proposed to improve the vehicle stability of unmanned vehicle under high-speed and low-attachment working conditions. Specifically, the path tracking controller is built by pre-targeting-following theory to achieve real-time tracking of the planned path of the unmanned vehicle; and the integrated control of the rear wheel turning angle and direct transverse moment is carried out by sliding mode control theory, and the torque distribution is realized by considering the front and rear axle loads and road adhesion coefficients to improve the body stability. The simulation results show that compared with the path tracking control strategy without considering stability, the path tracking strategy with considering stability decreases the average lateral tracking error by 5.2%, the peak mass lateral declination by 57.4% and the peak transverse pendulum velocity by 23% in the high-speed double shift line condition; in the low attachment mountain road condition, the average lateral tracking error decreases by 9.1%, the peak mass lateral declination by 54.3% and the peak transverse pendulum velocity by 1.4%. On the premise of ensuring the vehicle path tracking accuracy, the driving stability of the vehicle is improved. Reference | Related Articles | Metrics Select Correctness Detection of Smart Contract Based on Fuzzing WANG Jiacheng, JIANG Jiajia, ZHAO Jiahao, ZHANG Yushu, WANG Liangmin Computer Engineering and Applications    2024, 60 (5): 307-320.   DOI: 10.3778/j.issn.1002-8331.2211-0142 Abstract （38）      PDF（pc） （668KB）（34）       Knowledge map    Save The development of smart contracts is in its early stages. Different underlying programming languages and application platforms make the design of smart contracts lack specifications, which is prone to loopholes and losses. For the security vulnerability of smart contracts on Ethereum, it proposes a method for correctness detection of smart contracts based on fuzzing. This method generates fuzzy inputs based on the content and specifications of the smart contract, executes the smart contract in Ethereum virtual machine according to the fuzzy inputs, monitors the behavior of the contract in the execution process, generates multiple log files, extracts key information from the log files, triggers the test cases to get the vulnerabilities contained in the smart contract, and achieves the correctness detection. During the experiment, it detects 416 smart contracts for seven common vulnerability types and identifies 19 smart contracts as vulnerabilities. According to the analysis of artificial auditing, 18 of the 19 marked incorrect contracts do have security vulnerabilities. The experimental results show that the proposes method can identify the vulnerabilities contained in the smart contract with high accuracy, to detect the correctness of the smart contract. Reference | Related Articles | Metrics Select Improved YOLOv5 Model for Surface Defect Detection of Automotive Gear Components ZHU Deping, CHENG Guang, YAO Jingli Computer Engineering and Applications    2024, 60 (5): 321-327.   DOI: 10.3778/j.issn.1002-8331.2306-0425 Abstract （57）      PDF（pc） （633KB）（55）       Knowledge map    Save Aiming at the problems of low efficiency and poor precision in surface defect detection of automotive gear components, an improved defect detection method YOLO-CNF based on YOLOv5 is proposed. Firstly, add the CBAM attention module to the backbone network to make the model pay more attention to the defect areas of gear components and improve the ability to identify small defects. Secondly, the F2C module is designed to fuse shallow features, which alleviates the problem of the loss of small defect location information to a certain extent. Finally, NWD is used to optimize the regression loss to reduce the sensitivity to small target position deviations, and further improving the accuracy and precision of target positions. The experimental results show that the average precision of the improved algorithm reaches 86.7%, which is 3.2 percentage points higher than the original algorithm, and the detection speed is 43 frames per second. The improved algorithm basically meets the needs of the surface defect detection of automotive gear components. Reference | Related Articles | Metrics Select Improved Miner Chin Strap Detection and Personnel Tracking with YOLOv8s and DeepSORT DING Ling, MIAO Xiaoran, HU Jianfeng, ZHAO Zuopeng, ZHANG Xinjian Computer Engineering and Applications    2024, 60 (5): 328-335.   DOI: 10.3778/j.issn.1002-8331.2310-0280 Abstract （53）      PDF（pc） （695KB）（57）       Knowledge map    Save Ensuring proper safety helmet usage is of utmost importance in underground mining inspections to protect workers. However, challenging conditions, such as high temperatures, often lead to non-compliant helmet wearing behavior. Existing detection methods are insufficient for underground environments, resulting in low recognition accuracy and inadequate detection of correctly worn helmets. To address these issues, this paper proposes an improved version of the CM-YOLOv8s algorithm that focuses on the chin strap as a small target for safety helmet detection and compliance assessment. The DeepSORT algorithm is then employed to track workers who fail to comply with helmet-wearing regulations. To begin, a comprehensive dataset is curated utilizing underground surveillance cameras. The CM-YOLOv8s algorithm is leveraged for safety helmet detection by incorporating higher-resolution feature maps and introducing a cascaded query mechanism. This approach enables precise detection of small targets without significantly increasing computational costs. Furthermore, the enhanced DeepSORT algorithm is employed for person tracking by replacing the small residual network in DeepSORT with deeper convolutional layers, thereby enhancing the extraction of appearance information. The proposed algorithm is validated using a self-made dataset for underground safety helmet detection and tracking. Experimental results demonstrate that CM-YOLOv8s achieves an average precision of 92.3% for safety helmet recognition, which is a 4.2 percentage points improvement over YOLOv8s. Additionally, the average accuracy of the safety helmet compliance recognition system, based on CM-YOLOv8s and DeepSORT, is 85.37%, with a detection speed of 59 FPS. The proposed algorithm effectively addresses compliance detection in safety helmet wearing by accurately assessing the position of the chin strap in proximity to the individual’s jaw. It strikes an optimal balance between detection speed and accuracy while exhibiting robust adaptability to the complex underground environments. The successful implementation of this algorithm at the Chensilou Coal Mine over an extended period has demonstrated its efficacy in monitoring and providing early warnings for abnormal safety helmet wearing, thereby bolstering regulatory oversight and promoting the compliant use of safety helmets among miners. The algorithm holds great potential for enhancing safety measures in underground mining inspections and can be applied to similar industrial scenarios. Further research and development in this direction are warranted to expand its applicability and impact. Reference | Related Articles | Metrics Select CIEFRNet：Abandoned Objects Detection Algorithm for Highway LI Xu, SONG Huansheng, SHI Qin, ZHANG Zhaoyang, LIU Zedong, SUN Shijie Computer Engineering and Applications    2024, 60 (5): 336-346.   DOI: 10.3778/j.issn.1002-8331.2306-0395 Abstract （57）      PDF（pc） （928KB）（42）       Knowledge map    Save Highway abandoned objects endanger traffic safety, easily cause traffic accidents, so it is critical to recognize and clean them up in time. Due to the small area of highway abandoned objects in the image and complex image background, the existing detection methods often have the problems of missed and false detection. To address the above problems, an abandoned objects detection algorithm based on contextual information enhancement and feature refinement is proposed, which is called CIEFRNet. Firstly, a backbone feature extraction module (CSP-COT) incorporating contextual Transformer is designed to fully mine local static and global dynamic context information, and enhance the feature representation of small abandoned objects. In addition, the proposes improved spatial pyramid pooling (ISPP) is used in the backbone, multi-scale downsampling of features is realized by cascade dilated convolution, which reduces the loss of object detail information; in order to improve the feature fusion ability, a feature refine module (CNAB) is designed, in which a proposed mixed attention mechanism (ECSA) is embedded, which can suppress image background noise, and enhances the features of tiny abandoned objects. Finally, it uses the WIoU loss function based on dynamic non-monotonic focus mechanism to improve the learning ability of small abandoned objects and accelerate the network convergence. The experi- mental results demonstrate that the proposed method achieves 96.5%, 81.6%, 88.1% and 46.5% of accuracy, recall, AP0.5 and AP0.5：0.95 on the self-made highway abandoned objects dataset, respectively, which is better than the currently prevailing object detection methods, and its algorithm complexity is also lower to meet the needs of practical scene applications. Reference | Related Articles | Metrics Select Multi-Resource Cooperative Optimization of Container Terminals Considering Inter-Ship Transshipment Service WANG Xiaoguang, WANG Hongyu, LIU Boyu Computer Engineering and Applications    2024, 60 (5): 347-356.   DOI: 10.3778/j.issn.1002-8331.2210-0107 Abstract （44）      PDF（pc） （606KB）（40）       Knowledge map    Save To make efficient use of terminal resources and improve the operation efficiency of container terminals, based on the background of large-scale transshipment ports, aiming at the tactical decision-making problem of terminal resource allocation, considering the interaction of ship schedule stability, inter-ship container transshipment flow, berth scheduling and yard allocation, and based on the terminal active management strategy, a terminal multi-resource collaborative optimization model is established with the objective of minimizing the completion time of all container transshipment tasks. Based on this, an artificial immune algorithm based on tabu search is designed to solve the model. The effectiveness of the model and algorithm is verified by six numerical experiments of different scales and the comparison experiments with genetic algorithm and artificial immune algorithm. The comprehensiveness and superiority of the collaborative scheduling model are tested by case comparison experiments in different scenarios. The results show that the minimum transit time can be shortened by 4.33% and the maximum can be shortened by 20.19%. It provides an effective decision-making idea for the design of multi-resource allocation template of container terminal. Reference | Related Articles | Metrics Select Multi-Objective Shop Floor Scheduling Combining NSGA-II and CSA YANG Qing, XI Zhenzhen, GE Liang, LIN Xingyu, XING Zhichao Computer Engineering and Applications    2024, 60 (4): 315-323.   DOI: 10.3778/j.issn.1002-8331.2210-0486 Abstract （43）      PDF（pc） （2494KB）（29）       Knowledge map    Save Aiming at the simultaneous scheduling problem of scheduling jobs and automated guide vehicles (AGVs) in the flexible workshop system, consider building the objective function to minimize maximum processing machine duration, single AGV handling time, and total AGV handling time in the case of a finite number of AGVs and processing machines. Design an improved algorithm that combines NSGA-II (non-dominated sorting genetic algorithms) and clonal selection algorithm (CSA) to solve such problems. Firstly, the workpiece, the processing machine and the AGV are used for three-part coding. Secondly, the total score of non-dominated ranking and objective function value size sorting is introduced to stratify the population, so as to effectively retain excellent individuals. Thirdly, for the cloned population, different probabilities of variation are adopted for different levels of populations, and the genetic recombination of internal exchange and uniform cross-mixing exchange of chromosomes is carried out to effectively improve the diversity of the population and prevent it from falling into local optimum. Finally, three sets of comparative experiments verify that the algorithm has the advantages of short running time, high stability and good convergence when exploring the optimal solution. Reference | Related Articles | Metrics Select Predication Method of Continuous Non-Invasive Arterial Blood Pressure Using Fusion U-net Model WANG Jun’ang, ZHANG Lixin, WANG Sai, WU Kaifeng, KAN Xi, CHEN Naiyuan Computer Engineering and Applications    2024, 60 (4): 324-330.   DOI: 10.3778/j.issn.1002-8331.2211-0031 Abstract （57）      PDF（pc） （2180KB）（48）       Knowledge map    Save Continuous blood pressure monitoring is helpful to the diagnosis and treatment of cardiovascular diseases. At present, machine learning and deep learning are used to predict blood pressure by manually extracting feature parameters. This method cannot reconstruct complete blood pressure signals. Therefore, a continuous non-invasive arterial blood pressure measurement method based on the fused U-net model is proposed. Firstly, the original photoplethysmogram (PPG) signal is used as the input to reduce the error of manually extracting feature parameters. Secondly, the U-net network is used to reconstruct the arterial blood pressure signal. In order to further improve the accuracy of the predicted blood pressure waveform, the reconstructed blood pressure signal is used as the input of the MultiResUnet network. The MultiRes module is used to learn different features from the data. The Res Path module alleviates the semantic differences between the encoder and the decoder, making the model learning easier. The arterial blood pressure (ABP) waveform predicted by the fused U-net network in the subject evaluation of MIMIC -Ⅲ dataset is highly correlated with the actual waveform. The calculated mean absolute errors of systolic blood pressure (SBP), diastolic blood pressure (DBP) and mean pressure (MAP) are 2.20 ± 4.30 mmHg, 1.82 ± 3.146 mmHg and 2.25 ± 2.86 mmHg. The method satisfies the requirements of the Association for the Advancement of Medical Instrumentation (AAMI) standard and reaches Grade A in the British High Pressure Society (BHS) standard. Reference | Related Articles | Metrics Select Dialogue Symptom Inference Based on Structured Self-Attention Network PAN Dinghao, YANG Zhihao, LIN Hongfei, WANG Jian Computer Engineering and Applications    2024, 60 (4): 331-337.   DOI: 10.3778/j.issn.1002-8331.2211-0071 Abstract （27）      PDF（pc） （2099KB）（14）       Knowledge map    Save Symptom inference is a critical component of the medical dialogue system for automatic diagnosis. With the improvement of online diagnosis in recent years, the number of doctor-patient dialogue texts has continued to increase. The initial research on symptom inference based on electronic health records has gradually shifted to doctor-patient dialogue texts. However, most existing studies ignore the special role and symptom entity structure prior knowledge in dialogue, which can help the model learn contextual associations better. Therefore, this paper proposes an improved self-attention network based on the prior knowledge of role and entity structure and combines it with a pre-trained language model. The proposed model integrates the prior knowledge of role and entity structure into the encoding stage of the text, and can more accurately infer the attributes of symptom entities. Experimental results on the CHIP-MDCFNPC dataset of the CBLUE2.0 show that the proposed model outperforms the baseline model, which verifies the effectiveness of prior knowledge and model structure. Reference | Related Articles | Metrics Select Financial Fraud Recognition Method for Listed Companies Based on Deep Learning and Textual Emotion CAO Ce, CHEN Yan, ZHOU Lanjiang Computer Engineering and Applications    2024, 60 (4): 338-346.   DOI: 10.3778/j.issn.1002-8331.2305-0281 Abstract （65）      PDF（pc） （2187KB）（46）       Knowledge map    Save Financial fraud of listed companies refers to the untrustworthy behavior of distorting accounting information by improper means, which has a negative impact on company operations, economic development, and social interests. At present, more research focuses on financial digital data, and less research on text information and deep learning algorithms. Therefore, a financial fraud recognition method for listed companies based on deep learning and textual emotional feature is proposed. Firstly, the method selects and preprocesses the financial indicators, and uses Bi-LSTM to extract the emotional features of the stock review text. Then, the method uses the RCC (residual-cross-convolutional) parallel network to recognize financial fraud. The network uses residual network, cross network, convolutional network and long short-term memory network to extract financial fraud features in parallel, and uses batch normalization and full connection to obtain the final recognition result. The experiment results show that this method achieves better results than other models in recognizing financial fraud for listed companies, with a recall rate and AUC of 88.46% and 82.06% respectively. Reference | Related Articles | Metrics Select End-to-End Robotic Arm Vision Servo Research Combined with Bottleneck Attention Mechanism LIU Bingkun, PI Jiatian, XU Jin Computer Engineering and Applications    2024, 60 (4): 347-354.   DOI: 10.3778/j.issn.1002-8331.2207-0459 Abstract （28）      PDF（pc） （2333KB）（19）       Knowledge map    Save Aiming at the cumbersome feature extraction steps and poor real-time performance of traditional visual servoing algorithms, an end-to-end direct visual servoing algorithm based on convolutional neural network is proposed. By directly predicting the instantaneous linear velocity and instantaneous angular velocity of the camera installed at the end of the robotic arm, servo positioning works without the need to label handcrafted features, camera intrinsics, or depth information. Firstly, the image observed by the camera is input into the GhostNet feature extraction network for information extraction, and the bottleneck attention module (BAM) is integrated into the network to enhance the spatial information and channel information of the target object. Then, the fully connected layer is used as the speed regression function, the linear velocity and angular velocity are decoupled and regressed. Finally, a real-time capture method is used to create the dataset required for training in the real environment, and the velocity labels are generated according to the position-based visual servo control law algorithm. Efficient and accurate localization and tracking tasks are accomplished under the trained initial pose. The experimental test results in a large number of real scenes verify the effectiveness of the algorithm, and it is also robust to scene background information. Reference | Related Articles | Metrics Select Solving Time Dependent Vehicle Routing Problem Based on Improved Aquila Optimizer Algorithm SHI Xiaojuan, ZHAO Xingfang, YAN Long, TANG Yuan, ZHAO Huimin Computer Engineering and Applications    2024, 60 (4): 355-365.   DOI: 10.3778/j.issn.1002-8331.2211-0048 Abstract （68）      PDF（pc） （2678KB）（42）       Knowledge map    Save An improved Aquila optimizer (IAO) algorithm is proposed for the time dependent vehicle routing problem with time-varying speed, and the travel time calculation method of the time dependent road network is analyzed. An Aquila-customer (A-C) coding and decoding method is designed based on the characteristics of vehicle routing problem. Combining the four hunting methods of Aquila: expanding exploration, narrowing exploration range, expanding exploitation range and narrowing exploitation range, its intelligent search behavior is redefined, adaptive large neighborhood search (ALNS) strategy is introduced, and various neighborhood destroy operators and repair operators are designed. The inferior solution acceptance criterion is added to the algorithm, and two stagnation perturbation strategies, circular heuristic perturbation mechanism and elite perturbation mechanism, are proposed. Solomon benchmark and simulations based on Figliozzi test cases with genetic algorithms, particle swarm algorithms and ant colony algorithms demonstrate the optimization performance of the IAO algorithm, while the experimental results of real cases verify the superiority of the IAO algorithm in terms of convergence speed and solution quality. It is shown that IAO algorithm has the application value for solving time dependent vehicle routing problem. Reference | Related Articles | Metrics Select Path Planning of Robotic Arm Based on Improved RRT Algorithm Combined with A* LONG Houyun, LI Guang, TAN Xinxing, XUE Chenkang, YI Jing Computer Engineering and Applications    2024, 60 (4): 366-374.   DOI: 10.3778/j.issn.1002-8331.2303-0157 Abstract （64）      PDF（pc） （2566KB）（48）       Knowledge map    Save For the problem that the RRT (rapidly-exploring random tree) path planning algorithm generates a huge number of nodes when planning the obstacle avoidance path of robotic arm in high-dimensional space, resulting in a large burden of algorithm operation, poor obstacle avoidance performance, and easy to fall into local extremes, an improved RRT algorithm combining A* judgment function is proposed. The sampling method of RRT is changed to generate a sequence of multiple randomly sampled points each time, and the improved A* judgment function is used for sorting. Distance judgment is performed on each generated node to prevent it from falling into local search. Finally, a repetitive greedy strategy is used to remove redundant nodes, and cubic B-spline is used to make path smooth. The performance of the algorithm is analyzed in 2D and 3D maps and robot arm simulations and prototype experiments. The improved RRT algorithm can effectively reduce the number of nodes for the robotic arm to reach the target poses, alleviate the local extremes, and avoid obstacles to reach the target poses quickly and stably, which proves the effectiveness and superiority of the improved RRT algorithm. Reference | Related Articles | Metrics Select Optimization of Location-Routing for Collection Center Under Smart Waste Collection MA Yanfang, JIA Jiapeng, LI Zongmin, YAN Fang Computer Engineering and Applications    2024, 60 (3): 309-320.   DOI: 10.3778/j.issn.1002-8331.2210-0197 Abstract （40）      PDF（pc） （794KB）（47）       Knowledge map    Save With the rapid development of Internet of things technology and the improvement of public awareness of environmental protection, smart dustbins are gradually popular, making the waste recycling work face new challenges. Aiming at the location-routing problem with capacity constraints, a multi-agent optimization model of waste collection location-routing is constructed by introducing the two-commodity flow formulation. Smart recycling enterprise locates the collection center and aims to minimize the total cost in the upper level. Outsourcing transportation company selects the smart bins to be visited based on the recycling threshold, plans the recycling path and ensures that transportation costs are minimized in the lower level. An improved genetic algorithm is used to solve the problem: the upper layer uses clustering algorithm to determine the location of the collection center; the lower layer uses random generation and Clarke and Wright savings method to generate the initial population. And best-cost route crossover operator and inversion mutation operator are introduced. Based on Prins and Barreto benchmarks and compared with BKS, GAPSO and BSA, the average gap between the results and BKS is 0.419%. The smart waste collection routing method can effectively reduce the total cost, tested by simulating real cases, which provides decision support to solve the location-routing problem with capacity constraints in the context of smart waste collection. Reference | Related Articles | Metrics Select Clinical Feature Recalibration Attention Network for Cataract Recognition ZHANG Xiaoqing, XIAO Zunjie, ZHAO Yuhang, WU Xiao, Risa Higashita, LIU Jiang Computer Engineering and Applications    2024, 60 (3): 321-330.   DOI: 10.3778/j.issn.1002-8331.2210-0158 Abstract （31）      PDF（pc） （791KB）（24）       Knowledge map    Save In recent years, convolutional neural networks (CNNs) have been widely used for automatic age-related cataract classification. However, incorporating clinical prior knowledge of age-related cataracts into CNN design improves the classification performance and the interpretability of the decision-making process of CNNs, which has been less studied. To this problem, this paper proposes a clinical-feature recalibration attention network (CFANet) to classify age-related cataract severity levels automatically. In the CFANet, a simple yet effective clinical feature recalibration attention (CFA) block is designed to fuse clinical features adaptively by setting relative weights, aiming to highlight significant channels and suppress redundant ones. This paper conducts extensive experiments on a clinical AS-OCT image dataset of nuclear cataract and a public eye image dataset to verify the effectiveness of CFANet. The results show that CFANet outperforms advanced baselines by above 3.54 percentage points of accuracy on the clinical AS-OCT image dataset, such as squeeze-and-excitation network (SENet), efficient channel network (ECANet), style-based recalibration module (SRM). And the results on the public eye dataset also show that compared with strong attention-based CNNs and published works, proposed method obtains over 1 percentage point improvement. Moreover, this paper also uses visualization methods to analyze clinical feature weights and channel attention weights to enhance the interpretability of the decision-making process for proposed method. Reference | Related Articles | Metrics Select Dataset Enhancement Quality Evaluation Method for Chinese Error Correction Task as Example SONG Cheng, XIE Zhenping Computer Engineering and Applications    2024, 60 (3): 331-339.   DOI: 10.3778/j.issn.1002-8331.2210-0253 Abstract （31）      PDF（pc） （615KB）（19）       Knowledge map    Save Data augmentation is considered to be an effective solution to improve model performance. However, when selecting the generated data, it is necessary to consider the inherent data characteristics and the corresponding task relevance. Aiming at this problem, taking the Chinese error correction task scenario as an example, an evaluation method that can be used for dataset enhancement quality is proposed. The method uses the pre-training model optimized by contrastive learning to extract the feature vector of the dataset, and then proposes three basic evaluation indicators such as mutual coverage, total dispersion, and self-support, and gives a comprehensive dataset quality fusion indicator. The experimental analysis results on four data enhancement methods, two Chinese error correction data sets and three Chinese error correction models show that the above evaluation method can be independent of the test set performance inspection method, providing an important basis for the selection of different enhanced datasets. Reference | Related Articles | Metrics Select Defect Detection of Photovoltaic Modules Based on Multi-Scale Feature Fusion TIAN Hao, ZHOU Qiang, HE Chenlong Computer Engineering and Applications    2024, 60 (3): 340-347.   DOI: 10.3778/j.issn.1002-8331.2304-0390 Abstract （57）      PDF（pc） （661KB）（38）       Knowledge map    Save A photovoltaic modules defect detection algorithm based on multi-scale feature fusion is proposed to address the challenges of complex defect backgrounds, large differences in defect scales, and a high number of small target defects that traditional object detection algorithms cannot solve. The algorithm is based on the YOLOv5s framework. Firstly, a coordinate attention mechanism is embedded in the backbone network to extract important defect shapes and enhance the network’s feature extraction ability. Secondly, a bidirectional feature pyramid network is used in the Neck network to adaptively fuse image features of different scales using adaptive weights. Finally, a tiny target detection layer is added to the prediction layer, and the ASFF detection head is used to adaptively fuse different output layers to reduce the loss of target feature information. The improved algorithm is validated on a photovoltaic component dataset, and the experimental results show that it can quickly and accurately identify defects, with an mAP of 91.9% and a recall rate of 90.8%, which represents a 3.2 and 4.5 percentage points improvement in mAP and recall rate, respectively, compared to the YOLOv5s network. Reference | Related Articles | Metrics Select Application of ADASVM-CSLINEX Model Considering Misclassification Cost YANG Yuanyuan, LU Tongyu, CUI Jun, XU Wenfu Computer Engineering and Applications    2024, 60 (3): 348-356.   DOI: 10.3778/j.issn.1002-8331.2210-0379 Abstract （29）      PDF（pc） （659KB）（18）       Knowledge map    Save In binary classification prediction,there are two types of errors that inevitably occur in the prediction of stocks,but the misclassification costs of these two types of errors are often different in practical applications, and this article focuses on this issue. This paper introduces an asymmetric LINEX (Linear-exponential) loss function, which uses LINEX loss to achieve cost sensitive learning by penalizing the low cost of misclassification class linearly and punishing the high cost of misclassification class exponentially.The model uses SVM as the base classifier for AdaBoost, embedding LINEX loss function into AdaBoost-SVM weight update equation and updating the samples weights according to the different misclassification costs of positive and negative samples and whether the samples are misclassified or not. This paper takes the components of HS300 from January 2011 to December 2020 as samples for empirical research, and the proposed model is used to predict the rise and fall. The result shows that the ADASVM-CSLINEX model can obtain higher investment performance. Reference | Related Articles | Metrics Select Surface Defect Detection of Polarizer Based on Improved YOLOX-S Algorithm CHEN Le, ZHOU Yongxia, ZU Jiazhen Computer Engineering and Applications    2024, 60 (2): 295-303.   DOI: 10.3778/j.issn.1002-8331.2209-0375 Abstract （44）      PDF（pc） （711KB）（50）       Knowledge map    Save Polarizers are an essential part of liquid crystal displays, whose surface defects not only reduce the display quality of liquid crystal displays, but also cause the scrapping of the entire liquid crystal panel. Aiming at the problems that the surface defects of polarizers have large scale differences and various shapes, an improved YOLOX-S polarizer surface defect detection algorithm is proposed in this paper. The adaptive balanced feature pyramid (ABFP) module is proposed to sufficiently incorporate the multi-level features extracted by the backbone network, while increasing the detection branch through a single convolution to further enhance the multi-scale detection capability of the model. Then, an attention module CBAM is introduced in ABFP to focus on important features. In addition, the Mish activation function is used instead of the SiLU activation function while the CIoU loss function is adopted. The experimental results show that the improved algorithm achieves 92.97% and 55.16% of mAP50 and mAP50：95 on the polarizer surface defect dataset, which are 1.86 and 1.34 percentage points higher than that of YOLOX-S (FPN). Frames per second reaches 50, which basically meets the needs of industrial real-time detection. Reference | Related Articles | Metrics Select Weed Identification Method in Corn Fields Applied to Embedded Weeding Robots HE Quanling, YANG Jingwen, LIANG Jinxin, FU Leiyang, TENG Jie, LI Shaowen Computer Engineering and Applications    2024, 60 (2): 304-313.   DOI: 10.3778/j.issn.1002-8331.2211-0282 Abstract （61）      PDF（pc） （786KB）（47）       Knowledge map    Save In order to ensure the accuracy and rapidity of the embedded weeding robot in the corn field, a real-time target detection algorithm based on GBC-Yolov5s is proposed. First, the combination of the 1×1 convolution and depth-separable convolution is used to replace the traditional convolution, which reduces the redundant features generated by the backbone network without changing the size of the output feature map. Secondly, a bidirectional feature fusion network (S-BiFPN) network is designed to enhance the ability of feature extraction, which can make full use of different scale features to improve the speed of weed detection and combine the multi-channel structure with the self-attention mechanism to enhance the attention of small targets by compressing and reweighting the input features. Finally, MWeed data sets are built for different environments to test the proposed algorithm. The results show that compared with the Yolov5s and Faster RCNN model algorithms, the size of the GBC-Yolov5s model after lightweight is only 3.3 MB, the detection time of the input image (GPU) reaches 15.6 ms, and the average accuracy (mAP) reaches 96.3%, which can effectively improve the target detection speed and recognition accuracy, and provide a theoretical basis for the field of intelligent agricultural weeding. Reference | Related Articles | Metrics Select Optimization of Container Multimodal Transport Network Based on Underground Logistics System HOU Yujie, LIANG Chengji Computer Engineering and Applications    2024, 60 (2): 314-325.   DOI: 10.3778/j.issn.1002-8331.2212-0379 Abstract （52）      PDF（pc） （947KB）（46）       Knowledge map    Save In order to reduce the adverse impact of road transportation on urban traffic and environment, a new sustainable transportation mode——underground logistics system is gradually proposed to solve the common problems faced by the current development of port cities. Aiming at the layout scheme of underground logistics system in the Yangtze River Delta city agglomeration, a comprehensive transportation network optimization model is established from three perspectives：carbon emissions, time and cost, so as to analyze the rationality of implementing underground logistics system in port city agglomeration. The NSGA-III algorithm is used to solve the results that the introduction of underground logistics can reduce costs, and achieve the purpose of energy saving, emission reduction and alleviation of traffic congestion. Furthermore, NSGA-III algorithm is based on NSGA-II to optimize the crowding ranking, by introducing widely distributed reference points to maintain the diversity of the population. However, NSGA-III always deals with conflicting goals by prioritizing satisfying constraints, thus neglecting to maintain population diversity. Aiming at the problem that the population is trapped in the local optimal solution of the high-dimensional objective space, an improved NSGA-III algorithm is proposed to help the population span the large and discrete infeasible region by simultaneously dealing with the optimization objective and constraints. For this problem, “minimizing the shipping costs” “minimizing the vehicle waiting time” and “minimizing the carbon emissions” are selected as the objective functions, and the improved dynamic constrained NSGA-III is used for simulation analysis. The optimization results of the improved NSGA-III are compared with NSGA-III to verify the effectiveness of the algorithm. At the same time, it is proved that it can be applied to the practical logistics scheduling scheme. Reference | Related Articles | Metrics Select Fuzzy Dispersion Entropy and Its Application HU Baohua, ZHU Zongjun, JIN Feixiang, LU Cuiping, XIU Lei, WANG Yong Computer Engineering and Applications    2024, 60 (2): 326-336.   DOI: 10.3778/j.issn.1002-8331.2208-0360 Abstract （55）      PDF（pc） （846KB）（23）       Knowledge map    Save Dispersion entropy (DispEn) is a new dynamic index to measure the degree of signal irregularity. Compared with sample entropy, dispersion entropy can detect both amplitude and frequency changes at the same time and shorten the calculation time greatly. However, DispEn is sensitive to parameter selection, especially the number of classes (quantization level). Since the dispersion entropy is set based on the round function (step function), in some cases a small change in signal amplitude due to noise will alter the quantization sequence, thus changing the entropy value. In order to solve these limitations, fuzzy dispersion entropy (FuzzyDispEn) is proposed by combining fuzzy membership function and dispersion entropy. In FuzzyDispEn, fuzzy membership between embedding vector and quantization level is realized based on Euclidean distance. The advantages of FuzzyDispEn are tested using different synthetic time series signals. The results show that compared with DispEn, FuzzyDispEn has lower sensitivity to signal length and parameter selection, and better anti-noise performance. FuzzyDispEn is also used in completely analysis of EEG and bearing signal. Experimental results show that FuzzyDispEn also performs better than DispEn in real physical signal analysis. The results show that fuzzy dispersion entropy can provide a new method for signal complexity measurement. Reference | Related Articles | Metrics page Page 1 of 9 Total 325 records First page Prev page Next page Last page