Research And Applications Of Multi-Target Tracking Algorithm For Millimeter-Wave Traffic Radar

With the increasing applications of radar in military and civil fields,target tracking has become an indispensable part of radar signal procession systems.Millimeter-wave radar is widely used in civil fields because of its high resolution,strong anti-jamming capability and all-weather operation ability.In the intelligent transportation system,as an important data,traffic flow is mainly obtained by video cameras.However,the detection results of the cameras are not satisfactory in weather conditions such as rainstorm,fog and haze,which could have large deviations.With the aid of the millimeter-wave radar and multi-target tracking technologies,high precision measurement of traffic flow can be acquired.Therefore,the multi-target tracking algorithms of millimeter-wave radar and its applications in traffic flow statistics are investigated in this dissertation.The research work mainly includes the following aspects:(1)The target motion model of millimeter-wave radar,the Kalman filter theory,and related extension algorithms are introduced,included the linear Kalman filter(LKF),the extended Kalman filter(EKF),the unscented Kalman filter(UKF),and the adaptive unscented Kalman filter(AUKF).Simulations and performance comparisons are carried out to illustrate the differences between these filters.The tracking gate and traditional radar data association algorithm are introduced in details.Firstly,nearest neighbor data association(NNDA)algorithm is introduced in this dissertation.Secondly,the probabilistic data association(PDA)algorithm.Then,simulations are conducted and results analyses are given.(2)The multi-target tracking algorithm of millimeter-wave radar is studied and the multi-target tracking system is built,included plots centroid,the management mechanism of track state,the track initiation,and the track maintenance.A key study was carried out on the track maintenance.First of all,we propose an improved nearest neighbor point-track association algorithm.Using the track status to set the association priority,the steady track is associated first,and then the other state track is associated.Secondly,with regard to the track breakage phenomenon,which is a common issue in the measured data,the track-track association algorithm is designed to improve tracking stability and verified with the measured data.In the improved nearest neighbor point-track association algorithm,the track breakage could still occur when the plots error is relatively large.Finally,the improved PDA algorithm is introduced to improve the tracking performance,and the system is verified by the radar measured data.From the processing result of measured data,it can be seen that stable tracking can be obtained with the aid of the multi-target tracking system in the traffic scene.(3)For the problem of low accuracy of traffic flow statistics in intelligent transportation,a new method of lane detection and traffic flow statistics is proposed by using multi-target tracking technology.According to the track length,it can be determined that whether it is a real target to count up the number of traffic flow for each lane.The performance of the multi-target tracking system and the accuracy of traffic flow statistics are verified by several road tests.From the results of road test,it can be seen that the error of traffic flow is only 50%,which can meet the requirements of practical applications.