基于随机集的RBPF多目标关联跟踪算法  

Rao-Blackwellized Particle Filter Based on Random Finite Sets Theory for Multi-Target Association and Tracking

作　　者：赵欣[1,2] 姬红兵[1] 杨柏胜[1] 

机构地区：[1]西安电子科技大学电子工程学院,陕西西安710071 [2]中国航天科技集团公司第四研究院401所,陕西西安710071

出　　处：《电子学报》2011年第3期505-510,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.60677040)

摘　　要：针对大量杂波环境下数量变化的纯角度多目标航迹关联跟踪问题,提出一种新的基于Rao-Blackwellized粒子采样(RBPF)航迹关联的高斯混合概率假设密度(GMPHD)滤波算法.算法首先利用GMPHD在每时刻对多个目标组成的随机集合进行估计;然后利用基于随机有限集的RBPF对GMPHD所得到的目标集合进行检测和关联,有效解决GMPHD算法中无法进行多目标航迹识别的弊端;最后通过对所有粒子的融合完成航迹区分和估计.实验结果表明,提出方法比起目前经典的随机集Label-PHD关联跟踪算法,可以更有效的对数量未知的多目标航迹进行区分和关联估计,同时算法的跟踪性能及稳定性要好于Label-PHD算法.Due to the difficulty in association and estimation of multi-target tracks in the presence of data association uncertainty,clutter,noise and miss-detection.In this paper,a novel data association probability hypothesis density（PHD） filter for multi-target tracking based on Rao-Blackwellized particle filter（RBPF） algorithm is proposed.Firstly,the Gaussian mixture probability hypothesis density（GMPHD） filter has been proposed to estimate the set of all targets at every time step.Secondly,the data-association functionalities of RBPF can be incorporated with the PHD filter to produce the track-valued estimates of individual targets.Simulation results show that the proposed algorithm is more robust and accurate than Label-PHD algorithm which is very prevalent in the PHD tracking domains,also the proposed algorithm can estimate and distinguish each target more effective.

关 键 词：RAO-BLACKWELLIZED粒子滤波 多目标跟踪 概率假设密度滤波 航迹关联 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构]