基于箱粒子滤波的鲁棒标签多伯努利跟踪算法

针对在未知杂波和检测概率的跟踪环境中标准的标签多伯努利(LMB)算法跟踪精度较低、粒子覆盖集过大致使复杂度较高的问题,引入区间分析技术,提出基于箱粒子滤波的鲁棒LMB跟踪算法。建立目标增广空间模型,基于箱粒子滤波方法,推导出有杂波状态标签和LMB元素标签的预测、更新方程,并用多目标箱粒子LMB滤波递推估计目标状态。仿真结果表明,当杂波和检测概率先验未知,与现有非标签、非鲁棒算法相比,所提算法可实现在低检测概率和高杂波强度环境下对目标的稳定跟踪,同时大幅度提高算法的运行效率。

Robust Labeled Multi-Bernoulli Tracking Algorithm Based on Box Particle Filtering

The standard labeled Bernoulli (LMB) filter cannot guarantee a higher tracking performance,and multitude number of particles leads to the longer operation time of algorithm under the conditions of unknown clutter and detection probability.A robust labeled multi-Bernoulli algorithm based on box particle filtering is proposed.An augmented state space model is established,and the prediction and update state recursion equations with clutter state labels and LMB element labels are derived based on box particle filtering.The state of multi-target is recursively estimated using LMB filter box particles.Simulation reveals that the proposed algorithm has a better performance in target tracking under the conditions of unknown clutter and detection probability,and dramatically reduces the computation time with higher tracking accuracy under the conditions of lower detection probability and higher clutter ratet compared with the conventional algorithm with non-label and non-robustness.