基于高斯混合PHD滤波的多目标状态提取方法

MULTI-TARGET STATE EXTRACTION BASED ON GAUSSIAN MIXTURE PHD FILTER

高斯混合概率假设密度滤波(GM-PHD)方法可有效解决线性高斯模型下的多目标跟踪问题,在估计目标个数的同时提取多目标状态。但当杂波浓度过高、目标过于密集时,GM-PHD的状态提取精度有所下降。针对GM-PHD滤波算法在复杂环境下性能下降的问题,提出一种改进的GM-PHD滤波多目标状态提取方法,通过修正高斯分量更新权值,强化合并规则,降低密集目标和杂波造成的干扰。仿真实验表明该方法能在不同杂波环境下提高多目标状态估计的准确度。

Gaussian mixture probability hypothesis density( GM-PHD) filter can effectively solve the problem of multi-target tracking under the condition of linear Gaussian model,while estimating the number of targets it also extracts the states of multi-target. The state extraction precision of GM-PHD filter will drop down when it comes to the situation of closely spaced targets and too high clutter rate. In light of the performance degradation of GM-PHD in complex environments,we proposed an improved multi-target state extraction method of GM-PHD filter. By modifying the update weight of Gaussian component and enhancing the merging criterion it reduces the interference caused by intensive targets and clutters. Simulation experimental results showed that the propose method is able to raise the precision of multi-target state estimation in different clutter environments.