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

Multi-source information fusion has been widely used in both military and civilian fields,and multi-target tracking technology as a key technology has received more and more attention and research.Random factors such as the number of targets changing at any time and the continuous interference of clutter become normal in the tracking process.Traditional multi-target tracking algorithms will have problems such as large calculations and reduced tracking accuracy.However,random finite set theory provides different angles and directions in this respect,which is worth further study.In the process of recursive filtering,the target state and observation are modeled as a finite set,which provides a new idea for radar multitarget tracking to adapt to the change in the number of targets and does not require data correlation.In this paper,multi-target tracking is used as the research background.Standard cardinalized probability hypothesis density(CPHD)filters and joint generalized labeled multi-bernoulli(Joint-GLMB)filters cannot accurately track the non-linear moving target,expands the two nonlinearly and provides a solution to the improvement of the calculation efficiency of the improved algorithm.The main research of this article is as follows:1.Improved cardinalized probability hypothesis density filtering algorithm.In view of the application of the standard potential probability hypothesis density filter in the multiobjective system with nonlinear moving targets,considering that the cubature kalman algorithm is suitable for high-dimensional non-linear environment,and can achieve the third-order approximation accuracy through the spherical radial integration criterion.This paper studies the process of CPHD filtering algorithm based on the gaussian mixture method of cubature kalman filter when the nonlinear moving target exists in the multi-target system,and proposes a cubature kalman cardinalized probability hypothesis density(CK-CPHD)filtering algorithm.Finally,the simulation experiment is used to verify the CK-CPHD filtering algorithm.By comparing the optimal subpattern assignment(OSPA)distance with the sequential monte carlo CPHD(SMC-CPHD)filter under the same conditions,it is concluded that the performance of CK-CPHD filtering algorithm is better than that of SMCCPHD filtering algorithm.The final results show that the CK-CPHD filtering algorithm is suitable for the multi-target system with high-dimensional strong non-linear moving targets,and has good numerical stability,can accurately track the non-linear moving multi-target,and can accurately estimate the number of targets.2.Improved joint generalized labeled multi-bernoulli filtering algorithm.In order to obtain the track information of the target accurately in the multi-target tracking in the environment of the existence of nonlinear moving target,the labeled random finite set theory is introduced,and the label is used to mark the identity information of the target.In this paper,we study the joint generalized labeled multi-bernoulli filtering algorithm using the cubature kalman filtering technology(CK-Joint-GLMB),which can effectively and accurately estimate the state,number and track of the target.Aiming at the problem that a large number of useless components are produced in the operation of CK-Joint-GLMB filtering algorithm,a kind of CK-Joint-GLMB filtering algorithm with threshold is proposed.Through the threshold in advance,a large number of update components with low weight will be produced by prediction components and measurements,which will not participate in the later update steps,so as to reduce the calculation amount of the algorithm without losing the effect of the algorithm.Finally,the simulation experiment is used to verify the algorithm.Compared with the CK-Joint-GLMB filter algorithm under the same conditions on the OSPA distance,the results show that the CK-Joint-GLMB filter algorithm with threshold is superior to the original algorithm in algorithm efficiency on the basis of ensuring the performance of the algorithm.The target information and the track information can be accurately obtained,and the accurate tracking of multiple targets and the correct estimation of the track information in the environment where the nonlinear moving target exists are realized.