Non-line-of-sight Target Detection And Localization With Automotive Millimeter Wave Radar

Non-line of sight(NLOS)targets detection refers to the detection,positioning and tracking technology of NLOS targets in building and road scenes based on the electromagnetic wave multipath echo signals.It will have important application prospects and research value in many fields such as automatic drive,collision avoidance detection and urban battles.This thesis focuses on the problem of NLOS targets detection in urban building corner and road scenes,and carries out the research on multipath echoes modeling,NLOS targets detection and positioning based on the automotive millimeter-wave radar.The main contents are as follows:1.According to the feature of the typical building corner scene and the road NLOS front-front vehicle detection scene,first,the multipath propagation models of detecting NLOS targets in the two scenes are analyzed respectively.Then,the echo models of two scenes are established under the support of the experimental results.Finally,based on the real data,the energy losses of detecting targets at the same distance between the line of sight(LOS)path and the NLOS path are compared.2.To obtain the location of the NLOS targets in the building corner scene,based on the stationary automotive millimeter-wave radar platform,the signal processing algorithm for NLOS target detection and positioning is researched.Firstly,the echo preprocessing method for NLOS targets detected by stationary radar platform is studied,and the target's distance detection is realized by incoherent accumulation and cell average-constant false alarm rate(CA-CFAR)method.Then,the NLOS target localization algorithm is proposed,which uses the minimum variance distortion response(MVDR)method and the Kalman filter to achieve the accurate localization of NLOS moving targets.3.To achieve the detection and localization of the NLOS vehicle in the actual road scene,based on the moving automotive millimeter-wave radar platform,the detection and positioning methods are researched which depend on the multipath echo model.Firstly,the echo preprocessing method by the moving radar platform is studied,and the singular value decomposition method is used to separate the range profile of the LOS vehicle and the NLOS vehicle.The Gauss-Laplace algorithm is utilized to realize the denoising of the range profile.Then,the CFAR detection is respectively used to obtain the range of the LOS vehicle and the NLOS vehicle.Finally,the localization algorithm is proposed,which combines the multi-channel phase combination method with the density clustering method to realize the accurate localization of the LOS vehicle and the NLOS vehicle.