鲁棒标签多伯努利机动目标跟踪算法

Robust labeled multi-Bernoulli filter for maneuvering target tracking

针对未知杂波和检测概率的跟踪环境下,标准的标签多伯努利（LMB）算法对机动目标跟踪性能较差等问题,提出鲁棒标签多伯努利机动目标跟踪算法（R-LMB）.首先建立真实目标、杂波与检测概率的增广空间模型,然后结合多模型（MM）系统,推导出基于蒙特卡罗（SMC）实现的带有状态标签和LMB元素标签的预测与更新方程.研究结果表明：在杂波和检测概率先验未知的情况下,所提出的算法可实现对目标数和目标状态的准确估计,同时在低检测概率和高杂波强度环境中仍可保证良好的多机动目标跟踪性能.

In unknown clutter and detection probability environment, the standard labeled multi-Bernoulli（LMB） filter in maneuvering target tracking cannot guarantee a good performance.Aiming at the problems,a robust labeled multi-Bernoulli（R-LMB）filter for maneuvering target tracking was proposed. Firstly,an augmented state space model with actual targets clutters and detection probability was established.Then,under multiple model（MM）,prediction and update state recursion with clutter state labels and LMB element labels based on sequential Monte Carlo（SMC） implementation was derived. Simulation results reveals that the proposed algorithm can guarantee estimation accuracy on targets number and states with the unknown clutter and detection probability priori,and has a better performance on tracking maneuvering targets with lower detection probability and higher clutter rate.