Prohibited Items Detection in Baggage Security Based on Improved YOLOv5
Zuoshuai Wang, Hongyi Zhang, Zhibin Lin, Xiangqiong Tan, Ben Zhou
Abstract
To resolve the problems of high item overlap and complex background in the security detection of X-ray luggage images, this paper proposes a network model based on improved YOLOv5. A transformer is used in the last convolution module of the Backbone network, which effectively improves the feature extraction ability of backbone, and introduces the global attention mechanism module in Head branch, which solves the problem of complex background in X-ray baggage image and inhibits the interference of background. Finally, the adaptive spatial feature fusion algorithm is adopted in the prediction part to improve the accurate prediction ability of the model. The mean Average Precision (mAP) of the algorithm on OPIXray data set reached 91%, and the detection speed reached 47 FPS.