PHD多目标跟踪算法及参数影响分析

PHD Multi-Target Tracking Algorithm and Parameter Influence Analysis

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.