Research On Detection Algorithm Of Prohibited Items In X-ray Image

In public transportation scenarios such as airports,railway stations,and subway stations,the detection of prohibited for luggage packages has always been an important means of ensuring safety.Therefore,how to achieve efficient and accurate intelligent security is always explored by people.In order to take into account the efficiency and accuracy of the security inspection work,based on the deep learning theory,this paper designs two high-performance X-ray image prohibited items detection algorithms,and the specific research results are as follows:1.For the case of missed detection and false detection in the detection of X-ray image prohibited items,a network model of X-ray image prohibited items detection based on improved Capsule network is proposed.The overall framework of the algorithm consists of capsule network,feature enhancement module and feature screening module.First of all,the Capsule network can detect the same prohibited items with different postures in the form of vector capsules,while the feature enhancement module can extract richer feature information from the obscured and background complex images,and the feature screening module can retain useful feature information for the network to filter useless feature information.Experimental results on the published dataset show that the algorithm achieves superior performance and is better than the comparison algorithm.2.The prohibited in X-ray image is easy to have occlusion and different size,and the same prohibited will present different posture.So a network model based on improved Yolov4 X-ray image prohibited items detection is proposed.The overall framework of the algorithm by CSPDarknet53 as the backbone network for extracting features,with the use of deformable convolutional module from the adaptation to the extraction of posture and size of different sizes of prohibited items features,in order to send into the detection head of the features at the same time contain high features of speech information and low-level features of the details of information,we use multi-scale fusion module,the fused features are more conducive to improving the detection performance of the network.Experimental results on the published dataset show that the accuracy of the detection of prohibited items in X-ray images and the accuracy of positioning are significantly improved compared with other comparison algorithms.Based on the theory of deep learning,this paper proposes two algorithms for detecting prohibited items in X-ray images.The detection accuracy and positioning accuracy of prohibited in X-ray images have been significantly improved,which has important research value in the field of prohibited security inspection.The thesis includes 31 figures,13 tables and 60 references.