一种改进的多传感器粒子PHD滤波近似算法  

Improved approximation of multisensor particle PHD filter

作　　者：欧阳成[1] 姬红兵[1] 杨金龙[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

出　　处：《系统工程与电子技术》2012年第1期50-55,共6页Systems Engineering and Electronics

基　　金：国家自然科学基金(60871074)资助课题

摘　　要：概率假设密度(probability hypothesis density,PHD)滤波是一种有效的多目标跟踪算法。传统的PHD滤波只适用于单传感器,多传感器PHD滤波虽然理论上可行,但计算复杂度过高,实际中只能对其进行近似处理。迭代更新近似算法虽然简单易行,但滤波结果与参与更新的传感器顺序有很大关系,而乘积形式的多传感器PHD滤波近似算法由于存在缩放比例失衡问题,无法应用于工程实际。针对以上问题,提出了一种改进算法,先采用乘积形式计算联合似然,再采用求和形式计算缩放比例。仿真结果表明,该算法能够有效解决缩放比例失衡问题,在滤波性能和目标数估计方面均优于传统的迭代更新近似算法,具有良好的工程应用前景。The probability hypothesis density（PHD） filter is an effective algorithm for multitarget tracking,and the conventional PHD filter is only suitable for a single-sensor system.Since the multisensor version of the PHD filter is possible but computationally intractable,some approximations are proposed in many practical applications.A heuristic approximation,named iterated-corrector approximation,is the default approach for multisensor problems.However,the order of the sensors for updating impacts the filter results seriously in this algorithm.Then,a multisensor PHD filter is proposed to solve this problem,however,there is a scale unbalance problem in its implementation.Aiming at above problems,an improved algorithm is proposed,which calculates the joint likelihood function in the product manner and the scale factor in the summation manner respectively.Simulation results show that the proposed algorithm can solve the scale unbalance problem effectively and has a better performance than the iterated-corrector approximation in terms of state filtering and target number estimation,which has a good application prospect.

关 键 词：多传感器 多目标跟踪 随机有限集 概率假设密度滤波 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构]