基于拟蒙特卡罗方法的概率假设密度多目标跟踪  

Probability hypothesis density filter based on quasi-Monte Carlo method for multiple target tracking

作　　者：张慧[1,2,3] 韩崇昭[1,2,3] 闫小喜[1,2,3] 

机构地区：[1]西安交通大学智能网络与网络安全教育部重点实验室,西安710049 [2]西安交通大学机械制造系统工程国家重点实验室,西安710049 [3]西安交通大学电信学院,西安710049

出　　处：《控制与决策》2012年第8期1221-1225,1230,共6页Control and Decision

基　　金：国家973计划项目(2007CB311006);国家自然科学基金创新研究群体项目(60921003);国家自然科学基金面上项目(60921003;61074176)

摘　　要：为了改善多目标跟踪问题中概率假设密度(PHD)滤波的估计精度,提出基于拟蒙特卡罗的PHD滤波算法.该算法利用低偏差点集在状态空间中分布均匀的特性,使得采样粒子集最大程度地相互远离,充分地描述多目标状态的后验概率密度,从而准确地利用带有相应权值的粒子集来计算多目标数目和各个目标状态的估计值.仿真实验表明了算法的有效性,且估计性能优于粒子PHD滤波算法.To improve the precision of probability hypothesis density（PHD） filter when dealing with the problem of multiple target tracking, an implementation method of PHD filter based on quasi-Monte Carlo is proposed. This PHD filter algorithm uses the property of more regularly distribution of low discrepancy points and makes the sampling particles away from each other. Thus it can more fully describe the posterior probability distribution function and more accurately compute the estimate value of the target number and the state of individual target according to the particles with corresponding weights. Simulation results show that the modified algorithm is effective, and the estimation accuracy is superior to particle PHD filter algorithm.

关 键 词：多目标跟踪 概率假设密度 拟蒙特卡罗方法 低偏差点集 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]