狄拉克加权和概率假设密度滤波器  

The Dirac Weighted-sum Probability Hypothesis Density Filter

作　　者：刘宗香[1] 李丽娟[1] 谢维信[1] 李良群[1] 

机构地区：[1]深圳大学ATR国防科技重点实验室,广东深圳518060

出　　处：《信号处理》2015年第5期505-513,共9页Journal of Signal Processing

基　　金：国家自然科学基金项目(61271107;61301074);深圳基础研究项目(JCYJ20140418095735618);国防预研基金项目(9140C800501140C80340)资助课题

摘　　要：为解决在存在杂波、过程噪声协方差未知、目标数未知和变化情况下的多目标跟踪问题,提出了一种适用于线性系统模型的狄拉克加权和概率假设密度滤波器。该滤波器将多目标的后验矩表征为狄拉克加权和的形式。类似于高斯混合PHD滤波器,该滤波器在递归过程中传递多目标的后验矩。不像高斯混合PHD滤波器用卡尔曼滤波器获取多目标的后验更新矩,该滤波器采用变系数α-β滤波器获取多目标的更新后验矩。同时,也提出了一种变系数α-β滤波器中参数α和β的确定方法。仿真实验结果表明,所提出的滤波器为存在杂波、过程噪声协方差未知、目标数未知和变化情况下的多目标跟踪问题提供了一种有效途径,它的平均执行时间小于高斯混合PHD滤波器的平均执行时间,所以具有良好的工程应用前景。To solve the problem for multi-target tracking in the presence of clutter,an unknown covariance of the process noise,and an unknown and variable number of targets,we propose a Dirac weighted-sum probability hypothesis density（ PHD） filter for a linear system model. The proposed filter expresses the posterior intensity as the weighted sum of Dirac delta functions. Similar to the Gaussian mixture PHD filter,this filter propagates the posterior intensity of multiple targets in filter recursion. Unlike the Gaussian mixture PHD filter that uses the Kalman filter to obtain the updated posterior intensity of multiple targets,this filter employs α-β filter with variable gain to obtain the updated posterior intensity of multiple targets. Meanwhile,we also present a method for determining parameters α and β in the α-β filter with variable gain. The simulation results demonstrate that the proposed filter provides an efficient method for the multi-target tracking problem in the presence of clutter,an unknown covariance of the process noise,and an unknown and variable number of targets,and its average performing time is smaller than that of the Gaussian mixture PHD filter,so that it will have good engineering application prospects.

关 键 词：多目标跟踪 概率假设密度滤波器 狄拉克函数 Α-Β滤波器 线性系统 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN953[自动化与计算机技术—计算机科学与技术]