基于盒粒子滤波的多扩展目标伯努利滤波算法

MeMBer Filter for Multiple Extended Object Tracking Using Box Particle

针对非线性和区间量测条件下的多扩展目标跟踪问题,提出了一种基于盒粒子滤波的多扩展目标MeMBer滤波算法。该算法首先将盒粒子滤波引入到改进的ET-MeMber滤波中,并推导了适用于区间量测的多扩展目标伪似然函数和状态更新方程,接着给出了基于MD-AP聚类的区间量测集划分方法来处理滤波过程中存在的区间量测集划分问题。仿真实验结果表明,所提算法能够快速高效地对区间量测进行处理,与传统的序贯蒙特卡洛实现方式相比,具有更好的跟踪性能。

Aiming at the problem of multi-extended target tracking under nonlinear and interval measurement conditions,a MeMBer (Multi-Target Multi-Bernoulli,MeMBer) filter based on box particle filter is proposed.Firstly,we introduce the box particle filter into the improved ET-MeMber filter.The pseudo-likelihood function and state update equation for interval measurement are derived.Then,the interval measurement set partition method based on MD-AP clustering is applied to deal with the interval measurement set division problem in the filtering process.The simulation results show that the proposed algorithm can process interval measurement quickly and efficiently,and has better tracking performance than the traditional sequential Monte Carlo implementation method.