A Study For Radar Multi-Targets Tracking Algorithm Based On Random Finite Set

With the development of sensor network and information technology,multi-target tracking technology is developing rapidly,and it becomes more and more important in military defense.In the modern battlefield,the number of targets increases rapidly,and it keeps changing during the tracking process,which leads to the increase of the calculation amount and the decrease of the tracking accuracy within the traditional multi-target tracking algorithm.In terms of the above problems,the Random Finite Set(RFS)theory is proposed.It represents the target state and measurements as a Random Finite Set,and performs recursive Bayesian filtering based on the set,so avoids the data association algorithm through considering the whole target states and measurements as one.This paper focuses on the multi-target tracking algorithms based on RFS theory.(1)Multi-target tracking algorithm analysis based on RFS theory.This paper introduces the principle of multi-target tracking,and focuses on the basic theory of RFS and its application in multi-target tracking.Multi-target Bayesian filter are discussed,and three implementation algorithms are analyzed: Probability Hypothesis Density filter(PHD),Coordinate Probability Hypothesis Density filter(CPHD)and Multi-target and Multi-Bernoulli(Me MBer).Finally verified the three algorithms by simulation in a general tracking scenario,the result shows the CPHD has the best performance,but also increase the algorithm complexity.(2)Study on the gaussian mixture interactive multiple model adaptive birth intensity function PHD filtering algorithm.The original gaussian mixture PHD filter needs to initialize the birth intensity function,however,for airborne radar,the location of new targets cannot be determined,so the new target intensity function needs to cover the whole monitoring area,and resulting in low computational efficiency.To solve this problem,the adaptive birth intensity gaussian mixture PHD(ATBI-GMPHD)filter is proposed,it add tags to the original target state to distinguish new targets and existed targets,and use the measurements to generate new birth intensity function,then predict and update the new targets and existing targets separately.At the same time,the gaussian mixture PHD filter cannot accurately track the maneuvering target with fixed motion model,so the interactive multiple model(IMM)is introduced,interactive multiple model gaussian mixture PHD(IMM-GMPHD)filter.It can improve the performance for the multiple maneuvering target tracking under a complex environment.Finally,simulation results show that the improved algorithm can improve the accuracy of the multi-target tracking system.(3)Study on radar multi-target track reconstruction algorithm.To solve the problem of difficulty in examine and evaluating the performance of multi-target tracks in complex scenarios,a forward-backward PHD smoother is proposed to reconstruct the track through batch processing.It can used to adjust the track results to improve the accuracy of the overall multi-target track,so as to improve the performance of the radar multi-target tracking system.Finally,the simulation results show that the improved algorithm can improve the accuracy of the multi-target tracking system.