目标起始和维持规则辅助的SMC-PHD滤波方法  

An Improved SMC-PHD Algorithm Assisted by Target Initiation and Continuity Rules

作　　者：吴伟[1] 黄冶[1] 

机构地区：[1]电子工程学院501教研室

出　　处：《现代雷达》2013年第9期24-29,35,共7页Modern Radar

基　　金：安徽省自然科学基金(090412067)

摘　　要：针对标准SMC-PHD滤波方法在密集杂波和漏警条件下容易出现的目标数错估和失跟问题,提出一种应用目标起始和维持规则辅助的SMC-PHD滤波方法。在R-SMC-PHD中,每个粒子除了状态矢量和权值外,还增加了两个辅助变量,一个用来表示粒子的存活年龄,一个用来表示最近N个周期粒子受到观测值更新的情况。目标起始规则规定只有年龄大于某个预设门限值的粒子方可用于估算目标数及目标状态,以此减弱目标数高估问题;目标维持规则规定只有满足"M/N"规则的粒子可继续存活,并用伪观测值对漏警目标实现状态更新,以此消除因漏警引起的目标失跟问题。实验结果表明,在恶劣的跟踪环境下文中所提方法能够更好地动态估算目标数和目标状态,具有较强的新目标起始和存活目标维持性能。For the standard sequential monte carlo probability hypothesis density （SMC-PHD）, some problems such as error cardi- nality estimation and miss tracking are presented due to the difficulty of dense clutters, miss-detection and data association uncer- tainty. To eliminate these problems, a rule based SMC-PHD （R-SMC-PHD） assisted by target initiation and track continuity rules is proposed in this paper. Besides target state and weight, every particle has two accessorial variables, one is used to denote the persistent time, and the other is used to denote updated state by measurements in the last samples. The target initiation rule regu- lates that only particles which persistent age is larger than one preset threshold can be used to estimate the cardinality and targets state; the track continuity rule regulate that only particles which satisfy the ＂M/N＂ rule can be persisted, and the loss detection target can be updated by the pseudo measurements. The simulation results show that the proposed algorithm can provide cardinality more correctly and maintain target tracks more accurately, even in the abominable tracking situation.

关 键 词：多目标跟踪 序贯蒙特卡罗 概率假设密度 目标起始 目标维持 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]