一种新的SMC-PHD滤波的多目标状态估计方法

A Novel Multi-Target State Estimation Method for SMC-PHD Filter

针对现有的应用于多目标跟踪概率假设密度粒子滤波器的目标状态估计方法不能很好地解决目标密度较高情况下的多目标状态估计问题,提出了一种新的基于粒子标签的多目标状态估计方法。该方法利用附加在每个粒子上的身份标签将粒子分为不同的粒子群,粒子群的个数与概率假设密度粒子滤波器的目标估计个数相同。随后根据粒子与最近量测的似然函数估计目标的运动状态,使得粒子概率假设密度滤波器在目标密集的情况下仍能准确地估计出目标状态。仿真试验表明,论文所提方法在目标密度较大情况下能够较好地估计出多目标状态,并提高了目标关联的准确性。

For the problem that current state estimation methods used in multi-target tracking Probability Hypothesis Density（PHD） particle filter exhibit poor performance in the scene of dense multi-target tracking,a novel multi-target state estimation algorithm is presented in this paper.The algorithm utilizes the labels which assign to every particle for classify all the particles into different clusters,makes the number of clusters equal to the estimated target number,and exploits the likelihood function between latest measurements and particles to estimate target＇s state.The simulation results indicate that the proposed method can accurately estimate target＇s state in dense multi-target tracking and improve the performance of estimate-to-track association.