多机动目标跟踪的IMM-GMPHD滤波算法.doc

文档介绍：多机动目标跟踪的IMM-GMPHD滤波算法
文章编号:1003-6199(2011)04-0089-06?

摘要:针对现有多机动目标跟踪算法精度低、计算量大、约束条件苛刻等问题,本文将高斯混合概率假设密度(Gaussian Mixture PHD,GM?PHD)滤波器和交互式多模型(Interacting Multiple Model,IMM)相结合,提出交互式多模型GM-PHD(Interacting Multiple Model GMPHD,IMM?GMPHD)滤波算法。算法不仅避免了多目标跟踪中的数据关联问题,而且在漏检、目标密集、目标机动、航迹交叉、目标数目未知的杂波环境下能够稳定、精确地估计目标数目和状态。100次蒙特卡洛(Monte Carlo,MC)仿真结果表明,IMM?GMPHD滤波器能在不增加额外计算负担的基础上,体现出较高的精确度和较强的鲁棒性。
?
关键词:多机动目标跟踪;高斯混合概率假设密度;交互式多模型;蒙特卡洛?
中图分类号: TN953;TP302 文献标识码:A
??
IMM?GMPHD Filter for Multiple Maneuvering Targets Tracking

??
YAN Kang, YAN Yu?de?
(Nanjing University of Science and Technology, Nanjing 210094, China)

Abstract:Considering the traditional data association algorithm of multiple maneuvering targets tracking being of hard constraint condition, lower estimated accuracy, and plexity. In this paper an interacting multiple?model (IMM) implementation of the GMPHD filter is proposed to estimate the states of multiple targets which may have different maneuverability in measurements system. The new algorithm avoids the difficult problem of data association. It is able to deal with clutter, miss detection, false alarm, dense, and cross targets tracking effectively. 100 Monte Carlo (MC) simulation results show that the proposed MM?GMPHD filter significantly outperforms in estimating the number and states of the multiple maneuvering targets. The proposed algorithm has higher tracking accuracy and more steady tracking performance.
?
Key words:multiple maneuvering targets tracking;gaussian mixture;probability hypothesis density;interacting multiple model;monte carlo
?
1 引言?
多目标跟踪技术在军事和民用领域都有着广泛的应用,如空防、精确制导,以及汽车防碰撞、机器人视觉等。在多目标跟踪场景中,目标的出现和消亡,目标间的交叉、编队、迂回等各种协同和非协同机动,还有大量欺骗性干扰和复杂电磁环境和强杂波背景的复合干扰都对目标跟踪带来大量的不确定性。?
随着多目标跟踪理论研究的日益深入,形成了一些成熟的多目标跟踪方法,概括来讲可分两大类[1]:①极大似然类滤波算法②贝叶斯类滤波算法。这两类方法都需要数据互联理论和滤波技术共同完成,目标数目的不确定性让传统的状态空间方法难以处理,且随着目标量测的增长,数据互联的计算复杂度也呈指数级增长。近年来,基于随机有限集(
Random Finite Set,RFS)[2]理论的多目标跟踪算法受到了国内外学者的广泛关注。Mahler提出概率假设密度(