Research On Extended Object Tracking Technology Based On Random Finite Sets

For the extended target tracking problem,the traditional multi-target tracking algorithm is very computationally intensive because it involves data association.The emergence of the target tracking algorithm under the framework of Random Finite Sets(RFS)avoids the data association link and the computational complexity is greatly reduced,which has become a research hotspot in recent years.In this thesis,under the framework of random finite sets,the related techniques of extended target tracking are studied.The main research results are as follows:Firstly,in this thesis,target tracking technology based on random finite sets theory is studied,target tracking model is established,key techniques of extended target tracking are introduced,and probabilistic Hypothesized Density(PHD)filtering technology is studied and simulated.Secondly,when the measurement densities of different extended targets vary greatly,the traditional extended target PHD(ET-PHD)filtering algorithm is easy to be misclassified in the measurement division stage and computationally heavy.To solve this problem,a SNN similarity partitioning algorithm based on dynamic mesh density is proposed,and clutter is filtered before measurement partitioning to reduce the amount of invalid computation.Compared with the traditional ET-PHD filtering algorithm,the tracking accuracy and real-time performance of the proposed algorithm are improved by simulation.Then,in the absence of strength information of new target,the traditional ET-PHD filtering algorithm has poor tracking effect on extended target.To solve this problem,an ET-PHD filtering algorithm based on measurement-driven adaptive strength of new target is proposed in this thesis.And the SNN similarity criterion is introduced in the measurement division stage.The simulation results show that the algorithm which proposed in this thesis not only has advantages in tracking accuracy,but also greatly improves the running time of the algorithm.Finally,an improved adaptive threshold GM-CPHD algorithm is proposed to solve the problem of complex computation of extended target Cardinalized Probability Density(CPHD)filtering algorithm in clutter environment.The maximum likelihood adaptive threshold is introduced in the algorithm,and only the measurements within the threshold are used to update the filter.Simulation results show that the algorithm which proposed in this thesis can reduce the computational time and guarantee the tracking performance in the case of unknown clutter rate.