Research On The Application Of Track Initiation Technology In Multi-target Detection And Tracking

The target tracking technique has been continuously improved with the development of science and technology.It has a widely application in both military and civilian fields and is of important research significance.The existing research shows that the multi-target Bayesian filter can effectively track the target,but it can not provide the state estimations of the target at the initial time,which leads to the degradation of the tracking accuracy in the subsequent time and the generation of some missed detections and false alarms.The track initiation is the primary problem in target tracking and is also the basis of target tracking.Based on the above reason,we investigate the target initiation technique and proposes the multi-target Bayesian filter with the track initiation for target tracking by applying the track initiation technique to the multi-target Bayesian filter.The main contents of the thesis are as follows:1)We first introduces the multi-target Bayesian filter based on the finite set statistics theory(FSST)and its two approximate methods: the probability hypothesis density(PHD)filter and marginal distribution multi-target Bayesian(MDMB)filter.Then we outline the two sequential processing techniques of track initiation method: the rule-based track initiation method and logic-based track initiation method.2)Based on the traditional PHD filter and MDMB filter,we propose the rule-based PHD(RB-PHD)filter,logic-based PHD(LB-PHD)filter and rule-based MDMB(RB-MDMB)filter,logic-based MDMB(LB-MDMB)filter for two dimensions to solve the target teacking problem in the two dimensional space.We then compare the proposed filters with the PHD filter and MDMB filter,respectively,to test their tracking performance through the simulation experiments.The experimental results show that the proposed filters have a better track initiation ability and provides a more accurate state estimation.3)To solve the multi-target tracking problem in the three dimensional space,we also propose the RB-PHD filter,LB-PHD filter,RB-MDMB filter and LB-MDMB filter for three dimensions by applying the rule-based track initiation method and the logic-based track initiation method to the PHD filter and the MDMB filter,respectively.The trackingperformance of the proposed filters is also tested and analyzed by comparing them with the PHD filter and the MDMB filter through simulation experiments.