Application Of Target Detection And Positioning In Vehicle Security Inspection Scenarios

The safety inspection to vehicles and personnel is conducted by manual way at traditional highway police checkpoints,which reduces the efficiency and increases the uncertainty of safety inspection.Therefore,an intelligent security inspection scheme is proposed.The realization of the intelligence of the security inspection system is to apply the target detection and positioning technology to vehicle security inspection scenarios,and finally a fully automatic intelligent security inspection and verification system without human involvement can be realized.The main research contents of this paper are as follows:1)Design of intelligent security inspection and verification system.First,the requirements definition and preliminary design is carried out in the intelligent security inspection and verification system.Then the system process is analyzed.The intelligent security inspection system is divided into three modules: vehicle intelligent guidance shunt,target detection and positioning and witness verification.According to the research on target detection and positioning technology,a solution is proposed for the intelligent guideance shunt and target detection and positioning module.Finally,the target detection technology is applied to the business scenario of the intelligent security inspection and verification system for analysis.2)Video-based moving target detection.The purpose of moving target detection is to realize the intelligent guidance shunt of vehicles in an intelligent security inspection and verification system.First,the background subtraction method of the Gaussian mixture model is suitable for vehicle security inspection scenes and has good detection effect by comparing with several mainstream algorithms of moving target detection.Then the adaptive Gaussian mixture model combined with HSV color space is obtained through the improvement of the Gaussian mixture model.The complexity of the algorithm is reduced by the improved model automatically assigning a different number of Gaussian components to each pixel,and the self-shadow of moving targets is detected and removed by HSV color space.Finally,the improved model is applied to different scenes for comparative test.Experimental results show that the improved model has high detection accuracy and fast detection speed.3)Image-based stationary target detection and positioning.The stationary target detection and positioning in the image is to realize the stationary vehicle target detection and positioning in the intelligent security inspection and verification system.Firstly,the inherent attribute of the vehicle target in the image is analyzed,and it is concluded that the detection and location on local features of the vehicle is carried out based on the spatial position,physical structure and color features.Then GrabCut and histogram are used to detect and locate the rearview mirrors with fixed space,Hough is used to detect and locate the wheels with fixed physical structure,and HSV color space is used to detect and locate the license plates with fixed color features.Finally,the local feature detection and positioning algorithm is fast and effective in comparison with the detection algorithm combining SIFT,BOW and SVM.Through the research of video-based moving target detection and image-based stationary target detection and positioning,the application of two methods to vehicle security inspection scene can provide technical support for intelligent guidance shunt and target detection and positioning of vehicles in intelligent security inspection and verification system.Testing with real data in reality,the target detection and positioning technology has a good effect on detection,which sets an example to the design of intelligent security inspection system.