Research On Detection And Tracking Algorithm Of Motion Targets For Passive Millimeter-Wave Image

Passive millimeter wave imaging system forms images by receiving the radiation of the millimeter wave in the scene according to different millimeter wave radiation intensity of different materials.The system has advantages of harmless to human body,ability to protect privacy and detect metal and non-metallic objects when applied in the field of security.However,due to limited engineering,millimeter-wave image has low signal to noise ratio,and the target may occurred deformation,scale changes,occlusion with the human body movement in the security environment which make target detection and tracking difficult.Therefore,study target detection and tracking algorithm to achieve target effective detection,accurate position,continuous and stable tracking in millimeter-wave image has a high academic value and practical value.Based on the analysis of passive millimeter wave imaging principle,this thesis has studied the target detection and tracking algorithm for passive millimeter wave image based on human,metal objects and background radiation characteristics.The main research contents are as follows:(1)In the actual imaging scene,the similar bright temperature of background and target caused small difference between the gray scale of the target and the background in the millimeter-wave image.Target detection algorithm based on region extraction is proposed.The algorithm normalize the columns of the human region in the image,and then the histogram correction is carried out on the enhanced image.The weighted maximum entropy threshold segmentation method is used to solve the segmentation threshold in the new histogram,and then image segmentation,body area extraction and target extraction is carried out.Finally,the simulation is carried out in the millimeter-wave image.The results show that the algorithm has high detection rate and low false alarm rate.(2)In view of the fact that the target becomes blurred with the human motion boundary in the actual scene,it is even submerged in the noise,the detection algorithm based on the region extraction is difficult to detect the target,and then the missed detection phenomenon occurs.The discriminate tracking algorithm may cause drift when tracking target of scale change,deformation,occlusion.The fast tracking algorithm based on cyclic matrix is proposed to solve the above problems.The algorithm uses the multi-channel gradient histogram to describe the target,constructs the characteristics of the cyclic sample,uses the cyclic matrix property,uses the fast Fourier transform and its inverse transformation to solve the linear classifier and all the sample feature response to quickly predict the target position.Simulation results show that the algorithm is robust and robust to noise.(3)In view of the security environment,the average gray value of the manned scene and the unmanned scene is different,and the human body may carry multiple targets,the multi-target detection and tracking algorithm of the millimeter wave image sequence is proposed.The algorithm firstly determines the starting frame of the detection by the average gray scale judgment,and then takes the maximum number of frames in the ten frames of the detected image as the actual target number,and then carries out tracking algorithm to predict the target location when the target can not be detected in the subsequent image frame.The simulation results show that the proposed algorithm can achieve effective and stable tracking of each target when the relative distance of each target is constant.