基于势概率假设密度平滑器的多目标跟踪算法

Multi-target tracking algorithm based on cardinalized probability hypothesis density smoother

针对概率假设密度(PHD)滤波在杂波环境下对机动多目标进行检测与跟踪时,易出现高阶势分布信息丢失,从而导致目标检测出现偏差的问题,提出一种将势概率假设密度(CPHD)滤波与平滑算法相结合的多目标跟踪算法。从CPHD的预测与更新步骤出发,结合后向平滑递归公式,推导CPHD平滑公式,并提出基于高斯混合实现的GM-CPHD平滑器。仿真实验表明,GM-CPHD平滑器的检测与跟踪性能优于未经平滑处理的CPHD滤波器。

When multiple maneuvering targets are estimated and tracked with probability hypothesis density（PHD）filter in clutter,it is easy to lose higher order cardinality information which will result in the estimation deviation of multi-target.A multi-target tracking algorithm combined cardinalized probability hypothesis density （CPHD）filter with the smoothing algo-rithm is proposed.The formula of CPHD smoothing is deduced reasonably according to the prediction,update steps of CPHD and combined with the backward smoothing recursion formula.In addition,the CPHD smoother based on Gaussian mixture is also proposed.Simulation results show that the proposed solution is better than the CPHD filter without smoot-hing.