基于分布式PMHT的多传感器多目标跟踪
Multi-sensor multi-target tracking based on distributed PMHT
姚思亦 1李万春 1高林 1张花国 1胡航玮2
作者信息
- 1. 电子科技大学信息与通信工程学院,四川成都 611731
- 2. 北京机电工程研究所,北京 100074
- 折叠
摘要
在目标跟踪领域,概率多假设跟踪(probability multiple hypothesis tracking,PMHT)算法作为一种批处理算法,计算量远远小于传统的多假设跟踪算法.当前,PMHT算法的应用受限于集中式处理,本文首先在传统算法的基础上对传感器网络下的算法似然进行了推导,得到多传感器算法下的关联后参数,接着基于共识性处理策略进行了混合共识,最后使用卡尔曼滤波完成了对目标参数的后验估计,使得PMHT算法能够被应用于不包含融合中心的全分布式传感器网络多目标跟踪.实验结果表明,在不同的杂波密度下,分布式PMHT在跟踪误差上相对于单传感器算法有着90％以上的改善效果,与集中式算法相比跟踪性能接近且运算速度更快.
Abstract
In the field of target tracking,probability multiple hypothesis tracking(PMHT)algorithm,as a batch processing algorithm,has much less computation than the traditional multiple hypothesis tracking algorithm.Currently,the application of PMHT algorithm is limited by centralized processing.On the basis of the traditional algorithm,this study firstly derives the algorithm likelihood under sensor network to obtain the post-correlation parameter under multi-sensor algorithm,followed by hybrid consensus based on the consensus processing strategy,and finally the posteriori estimation of the target parameters is accomplished by using Kalman filtering.This study enables the PMHT algorithm to be applied to the fully distributed sensor network without fusion centers.The experimental results show that under different clutter densities,the distributed PMHT has more than 90％improvement in tracking error compared to the single-sensor algorithm.Distributed PMHT has close tracking performance and faster computation compared to centralized algorithms.
关键词
多目标跟踪/概率多假设跟踪/一致性共识/集中式状态估计/分布式状态估计Key words
multi-target tracking/probability multiple hypothesis tracking(PMHT)/consensus/centralized state estimation/distributed state estimation引用本文复制引用
出版年
2024
NETL
NSTL
万方数据