Research On Multi-target Tracking Algorithm For Automobile Anti-collision Radar

The automotive radar system,an active collision prevention measure in intelligent traffic fields,plays a practical role in realizing self-driving car and reducing the incidence of traffic accidents.Based on the automotive radar system,the multi-target tracking algorithms estimate the number and state information of the detected vehicles reliably and timely and thereby enhancing the reliability in cluttered road environment.This dissertation focuses on multi-target tracking algorithms for automotive radar system in road environment.The main contents of the dissertation can be summarized as follows.Firstly,in order to tackle the problem of non-linear tracking for a time-varying number of maneuvering targets for the automotive radar,a novel DS-UKGMPHD algorithm is presented.Diagraph switching(DS)variable structure is used to accommodate various target motions in real world.In addition,Unscented Kalman(UK)filter and Gaussian mixture probability hypothesis density(GMPHD)filter are also used thereafter.Simulation results demonstrate the superiority of the presented algorithm to IMM-UKGMPHD filter in terms of estimation accuracy of both number and states.Secondly,in order to tackle the problem of tracking a varying number of maneuvering targets in cluttered road environment with unknown Target Birth Intensity,a novel algorithm based on detection of new targets is studied.The algorithm uses the multiple feasible hypotheses algorithm to detect the positions of the new targets and then uses DS-UKGMPHD filter to track vehicles.Simulation results indicate that the presented algorithm can be used to tackle the above-mentioned problem and thereby broadening the application scope.Thirdly,in order to tackle the problem of the targets' discarded information caused by radar misdetection,a novel algorithm,VSMM-IGMPHD,is studied.The algorithm uses DS algorithm to realize the variable structure.Additionally,the UK filter and improved GMPHD filter are combined as IGMPHD filter and used after DS algorithm.Simulation results show that the presented algorithm performs better in terms of both number and states of the road targets than VSMM-GMPHD when radar misdetection occurs.