Research on Dangerous Goods Detection Based on Improved YOLOv8s X-ray Images
KONG Zhicheng1 ,LIU Xingde2 ,XIE Yannan1 ,REN Luoying1,YU pengze1
1.School of Information and Control Engineering,Jilin University of Chemical Technology, Jilin City 132022,China；2.School of Mechanical and Electrical Engineering,Jilin University of Chemical Technology, Jilin City 132022,China
Abstract: Safety is the basis of human survival and development, especially in banks, railway stations, museums and other densely populated places, danger may occur at any time, aiming at the problems of missed detection and false detection of dangerous goods caused by the lack of attention of front-line security personnel, this paper proposes to improve the YOLOv8s on the detection of dangerous goods in X-ray images. Firstly, a dynamic convolution module (ODCov) is introduced on the basis of the original model to improve the accuracy of dangerous goods detection and feature extraction ability, especially in complex environment detection. Secondly, an efficient Multi-scale Feature Fusion Network (BiFPN) is introduced, which can better integrate features of different scales and improve the detection ability of the model for multi-scale targets. Finally, the Exponential Moving Average attention mechanism is introduced to improve the target detection performance and obtain more detailed information of dangerous goods. The improved YOLOv8s model was tested on the SIXray dataset, and the average accuracy of the test results reached 93%, which was 1.8% higher than that of the original model,Achieve an 81.7% increase in frames per second.
孔志成, 刘兴德, 谢延楠, 任洛莹, 余鹏泽. 基于改进YOLOv8sX光图像危险品检测研究[J]. 吉林化工学院学报, 2026, 43(1): 61-68.
KONG Zhicheng, LIU Xingde, XIE Yannan, REN Luoying, YU Pengze. Research on Dangerous Goods Detection Based on Improved YOLOv8s X-ray Images. Journal of Jilin Institute of Chemical Technology, 2026, 43(1): 61-68.
