Extended Targets Tracking For Through-the-Wall Imaging Radar

Through-the-wall imaging radar(TWIR)can detect the targets behind walls by passing electromagnetic waves with a low frequency band through building walls.It is suitable for military or civilian fields such as homeland security,natural disaster search and rescue,and crisis management.Due to the high-resolution of the through-the-wall imaging radar,the human target occupies multiple pixels on the image.The high maneuverability of the human body makes the shape and size of the image unstable.Moreover,the interaction between the targets causes the temporarily loss of the targets in the multi-target scenario.These problems will increase the difficulty of the human target tracking in the actual scenario.This thesis focuses on the robust extended targets tracking.In terms of imaging,image preprocessing methods and extended targets tracking after imaging,the contributions are listed as follows:1.To describe the human targets with multi-scattering,the human target is modeled as an extended target.The signal model for the TWIR is established.To suppress the influence of the wall,the data preprocessing methods based on first-order canceller and image self-focusing are presented.The signal-to-noise ratio(SNR)of the range profile is improved and the defocus problem is alleviated.To suppress the grating lobes of the multiple-input-multiple-output(MIMO)array,the imaging preprocessing methods based on energy coherence factor(ECF)and phase coherence factor(PCF)weighting and spatial filter are proposed,which can achieve the clean target images.The experiment verifies that the image preprocessing method based on PCF weighting has a better performance on grating lobes suppressing.The above research lays the foundation for the following methods of tracking after imaging.2.To deal with the problem of single extended target tracking,a tracking framework for single extended target in image domain is presented,which reduces the computational complexity of tracking after imaging.To improve the splitting of the extended target image,the mean-shift algorithm based on amplitude histogram is proposed which can mitigate the splitting and enhance the stability of extend target tracking.To improve the tracking accuracy of the target with high maneuverability,a particle filter based on amplitude histogram is proposed.The tracking failures caused by temporary loss is alleviated.The robustness of tracking human target with high maneuverability is improved.3.To describe the multiple extended targets tracking problem,combined with the tracking framework of single extended target in image domain,a tracking framework of multiple extended targets in image domain is proposed.To deal with the problem of the occlusion and the changing of targets' scale,the mean-shift algorithm is developed and combined with Kalman filter.A multi-target tracking based on multi-algorithm is proposed,which has a greater performance than tracking only by mean-shift algorithm when target occlusion happens.The robustness of tracking multiple human targets is enhanced.To suppress the problems that the azimuth resolution of portable TWIR is low and the overlap between targets,the oblique elliptical extended target model is analyzed.The extended-target joint probabilistic data association(ET-JPDA)algorithm is presented,which breaks the one-to-one correspondence hypothesis between measurement and target.Comparing with tracking in image domain,the auto-scaled tracking frame fits the contour of the target better and tracks more robustly when the target image overlaps.The above imaging and tracking algorithms have been verified by experiments and have been proven to be able to track single or multiple moving human targets in real scenarios.