基于概率假设密度平滑器的低检测概率下多目标跟踪

Multi-target Tracking under Low Detection Probability Based on Probability Hypothesis Density Smoother

针对传感器低检测概率下多目标跟踪的问题,提出了一种概率假设密度(PHD)滤波平滑器,并给出了该平滑器的高斯混合(GM)形式.综合运用前向PHD滤波递推与后向平滑两个步骤,改善了多目标跟踪系统对目标误跟踪的情况.通过仿真结果说明,经过平滑的PHD滤波与未经平滑的PHD滤波相比,在目标数目与状态的估计精度上得到了明显的提高.

To solve the problem of multi-target tracking under the condition of low detection probability of sensors, we propose a probability hypothesis density （PHD） smoother, and give the Ganssian mixture （GM） form of the smoother. The algorithm takes use of PHD forward recursion and backward smoothing, which lessens the possi- bility of wrong tracking of the target under the condition of low detection probability of sensors. In addition, the simulation results demonstrate that , when comparing the smoothed PHD filtering with the unsmoothed PHD fil- tering, the estimation accuracy of the number and condition of targets is significantly improved.