基于PMBM滤波的机动非椭圆扩展目标跟踪算法  

A Maneuvering Non-ellipsoidal Extended Target Tracking Algorithm Based on PMBM Filter

作　　者：李佳彤 杨金龙 LI Jiatong;YANG Jinlong(School of Artificial Intelligence and Computer Science,Jiangnan University,Wuxi,Jiangsu 214122,China)

机构地区：[1]江南大学人工智能与计算机学院,江苏无锡214122

出　　处：《信号处理》2023年第6期1108-1119,共12页Journal of Signal Processing

基　　金：江苏省自然科学基金项目(BK20181340);国家自然科学基金项目(61305017)。

摘　　要：扩展目标的形状大多是不规则的,且观测角度不同会发生变化,此外,在复杂场景下,还会存在扩展目标和点目标同时存在,且存在机动的问题,针对该问题,本文提出一种混合形状多模型泊松多伯努利混合滤波算法(IMM-VNNET-PMBM)。首先,通过子随机矩阵表示的形状多子目标非椭圆伽马高斯逆威沙特(GGIW),代替泊松多伯努利混合滤波框架下表示扩展目标的单GGIW密度,以提升形状估计精度;其次,目标运动状态发生变化时,扩展目标的形状信息通常也会发生变化,采用固定的形状子目标估计扩展目标形状将不准确,算法中通过基于密度的DBSCAN算法进行量测聚类,以多假设数据关联计算形状子目标分量的存在概率,自适应调整子目标数量,实现对扩展目标的时变形状估计;最后,算法通过融入交互多模型,实现了对复杂环境下的点目标和多机动扩展目标同时跟踪;实验结果表明,提出算法能够有效对机动且形状变化的多扩展目标跟踪,具有较好的跟踪性能。The shapes of extended targets are mostly irregular and changes with different observation angles.In addition,there is problem of tracking maneuvering targets in complex scenarios of coexisting point and extended targets.To solve these problems,a maneuvering non-ellipsoidal extended target tracking algorithm based on Poisson multi-Bernoulli mixture filter is proposed.Firstly,the multiple sub-objects gamma Gaussian inverse Wishart(GGIW)represented by a subrandom matrix is used to replace the single GGIW density representing the extended target in the Poisson multi-Bernoulli mixture(PMBM)filter for improving the shape estimation accuracy.Secondly,when the target motion state changes,the shape information of the extended target usually changes as well,and consequently,the performance of multiple target tracking algorithms based on the constant number of sub-objects will decrease.Thus,DBSCAN algorithm based on density is employed for measurement clustering,achieving time-varying shape estimation of extended target by calculating the existence probability of sub-target components,thereby adaptively adjust the number of sub-targets.Finally,the algorithm achieves simultaneous tracking for coexisting multiple point and maneuvering extended targets by integrating interactive multiple model algorithm.Experimental results show that the proposed algorithm can effectively track maneuvering multiple extended targets with a good tracking performance for shape changes.

关 键 词：泊松多伯努利混合滤波 伽马高斯逆威沙特分布 非椭圆扩展目标跟踪 交互式多模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]