Research On Multi-target Tracking Algorithm Based On Random Set Theory In Complex Environment

With the continuous progress of science and technology,the sensitivity and accuracy of various sensors have been greatly improved,so the development of observation-based multi-target tracking technology is imminent.Especially in the complex environment,the observation of highly maneuverable targets becomes more difficult,and the performance of traditional multi-target tracking algorithms based on data correlation technology has been difficult to achieve.Therefore,in recent years,FISST which developed from RFS theory has become a new research direction in the field of multi-target tracking.Since the random finite set theory operates directly on the set,there is no need for data association,which can effectively avoid the problem of difficult data association of highly maneuverable targets in complex environments.Therefore,based on the random finite set theory,this thesis conducts in-depth research on the SMC-PHD multi-target tracking algorithm,and proposes corresponding solutions to the problems in multi-target tracking in complex environments.The main research work of the thesis is as follows:(1)Aiming at the problem of target state mutation and particle Impoverishment in SMC-PHD filtering algorithm in complex environment,a novel ST-SMC-PHD multi-target tracking algorithm is designed.Firstly,for the particle filter algorithm to track the hidden state of the nonlinear state mutation system,due to the degradation of estimation accuracy caused by particle Impoverishment,an adaptive intelligent particle filter algorithm AIPF based on Student's t distribution is proposed.By adaptively generating large weight and small weight particle sets,and then performing adaptive weighted crossover and mutation operations on the particle set,the particle diversity is improved,and the accuracy and robustness of the particle filter algorithm are improved.Simulation experiments verify the effectiveness of the algorithm.Through the method of partition sampling,the AIPF particle filtering algorithm is introduced into SMC-PHD filtering,and the ST-SMC-PHD multi-target tracking algorithm is designed.Through analysis and experiment,the algorithm can effectively improve the performance of the resampled particle set in SMC-PHD,reduce the negative effects caused by particle Impoverishment,and improve the stability of the algorithm.It effectively improves the tracking performance of the algorithm in the high clutter environment for the cross trajectory target with sudden change of state and the sharp trajectory target.(2)Aiming at the problem of particle degeneracy and computational efficiency degradation of SMC-PHD filtering algorithm in complex environment,the KB-SMC-PHD multi-target tracking algorithm is designed and implemented.Firstly,based on KLD sampling and bat algorithm,an adaptive particle filter algorithm KBPF which can dynamically adjust particle size is proposed for particle filter due to particle degeneracy.KLD sampling is used to dynamically adjust particle size in importance sampling.Then,the bat algorithm is used to optimize the particle set,and the bat algorithm interacts with the KLD sample in the iterative update,so as to achieve the purpose of improving the calculation accuracy and computational efficiency.The feasibility and effectiveness of the algorithm are verified by experiments.The KBPF algorithm is introduced into SMC-PHD filtering,and the objective function and global optimization strategy are redesigned.The KB-SMC-PHD multi-target tracking algorithm is designed.The algorithm can adaptively drive the particles to the region optimal solution flight,improve the calculation accuracy,and adaptively delete redundant particles to improve the efficiency of the algorithm.Simulation results show that the algorithm can effectively improve the accuracy of the algorithm and the efficiency of the algorithm.(3)Aiming at the problem that the SMC-PHD algorithm makes a wrong judgment on the target number in the complex environment,which affects the performance of the K-means algorithm in state extraction,an AP-SMC-PHD filtering algorithm based on AP clustering is proposed.Because AP clustering does not need to predict the number of target classes a priori,and can give the characteristics of determining the centroid,it fundamentally avoids the negative impact on the state extraction due to the inaccurate estimation of the target quantity.Simulation experiments show that the proposed algorithm has a significantly improved estimation accuracy in high clutter environment.(4)The clustering algorithm in SMC-PHD is difficult to effectively solve the problem of MEE when the target trajectory intersects.A DYE-SMC-PHD dyeing filter algorithm is proposed.Firstly,the particles are labeled by the dyeing strategy,the particle sets are divided according to the particle chromaticity,and the multi-objective state extraction is realized by the weighted MEE method.Simulation experiments show that the proposed algorithm has better computational performance in high clutter environment and higher estimation accuracy in the region where the target trajectories intersect.In this thesis,we design ST-SMC-PHD algorithm,KB-SMC-PHD algorithm,AP-SMC-PHD algorithm and DYE-SMC-PHD algorithm for the problems of SMC-PHD multi-target tracking algorithm in complex environment.Problems are as follows.The degradation of particle filter performance leads to the problem of low computational accuracy.Due to the misdetection caused by high clutter,the number of targets is not accurate.Tracking accuracy is low when moving targets include complex motions with sharp turns and trajectory crossings.The effectiveness of the above algorithm is verified by simulation experiments.The research results are of great significance for effectively improving the performance of multi-target tracking algorithms in complex environments,and have broad application prospects in practical multi-target tracking problems.