PHD多目标跟踪算法及参数影响分析  

PHD Multi-Target Tracking Algorithm and Parameter Influence Analysis

作　　者：吉楠[1] 董福安[1] 杨珺[2] 高生强[1] 

机构地区：[1]空军工程大学理学院,西安710051 [2]西安理工大学自动化学院,西安710048

出　　处：《电光与控制》2009年第1期75-79,共5页Electronics Optics & Control

基　　金：国家自然科学基金资助项目(60573040)

摘　　要：多目标跟踪的关键就是对目标数和目标状态的准确估计。将目标集合看成一个随机集,并且目标数也是变化的。采用一阶统计矩近似表示状态空间的概率密度,通过蒙特卡罗模拟近似表示一阶统计矩,从而实现多目标跟踪。实验表明,在杂波环境下,PHD算法可以实现多目标跟踪,并且各参数对跟踪精度有一定的影响。Exact estimation of target quantity and target state is crucial for multi-target tracking. The target set was taken as a random set, in which the target quantity was variational. The probability density of state space was expressed approximately by the first-order statistical moment. Then Monte Carlo simulation was used to express the first-order statistical moment for implementing multi-target tracking. Experiments showed that PHD algorithm can be used to track multi-target in clutter, and some parameters have certain influences on tracking accuracy.

关 键 词：多目标跟踪 有限集统计 概率假设密度(PHD) 粒子滤波 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP274[自动化与计算机技术—检测技术与自动化装置]