基于熵惩罚的EM未知杂波估计的PHD多目标跟踪算法

PHD Multi-target Tracking Based on Entropy Penalized EM of Unknown Clutter Estimation

针对概率假设密度多目标跟踪算法中存在的杂波强度未知的问题,提出一种基于熵惩罚的EM未知杂波估计的PHD多目标跟踪(EPEM-PHD)算法。首先采用有限混合模型对未知杂波密度建模,其次分别对混合权重及缺失参数施加熵惩罚因子,然后通过自适应动态系数调节,使得混合模型低权值分量加速消亡,减少了算法迭代次数,且算法对初始参数不敏感。仿真结果表明,该算法在杂波强度未知的环境下,具有精度高、跟踪稳定的优势,提高了PHD滤波器在多目标跟踪中的性能。

Aiming at the unknown clutter intensity existed in Probability Hypothesis Density （PHD） multi- target tracking algorithm, a tracking algorithm based on Entropy Penalized Expectation Maximization （EPEM） of unknown clutter estimation is proposed, called EPEM-PHD. The clutter intensity is modeled by finite mixture model. The entropy penalized factor is applied on mixed weight and the missing parameter. By adjusting the adaptive coefficient, the extinction of components with low weight is accelerated, thus can decrease the times of iterations. And the algorithm is not sensitive to initial parameters. Simulation results show that： the algorithm has the advantages of high precision and stable tracking, which improves the performance of PHD filter in multi-target tracking.