迭代最小斜度单型Sigma采样UPF算法.docVIP

基于Meanshfit聚类---Bhattacharya观测似然度修正的JPDA改进算法 田隽1,2 邵晓根 2 厉丹1,2肖理庆2 （1.徐州工程学院 江苏省大型工程装备检测与控制重点建设实验室，江苏 徐州221008； 2.徐州工程学院 信电工程学院，江苏 徐州221008） 摘 要：为解决多目标航迹聚集时JPDA联合关联事件的计算复杂度问题，提出一种基于meanshift聚类-Bhattacharya（Bhy）观测似然度修正的JPDA改进算法。利用meanshift得到聚类中心，据聚类中心与目标预测量测马氏距离形成跟踪门；提出Bhy似然度矩阵，将meanshift聚类中心与各量测Bhy距离所表征的观测似然度作为确认矩阵小概率事件划分依据，消除确认矩阵中小概率事件对联合关联事件计算复杂度的影响。实验结果表明：多目标航迹聚集时，本文算法减少计算复杂度同时保持了较高关联精度，跟踪性能明显优于经典JPDA。 关键词：多目标航迹聚集；Meanshift聚类；跟踪门；Bhattacharya似然度矩阵；JPDA 中图分类号：TP391 文献标识码：A An Improved JPDA Algorithm Based on Meanshfit and Bhattacharya Likelihood Matrix （1.Jiangsu Key Laboratory of Large Engineering Equipment Detection and Control,Xuzhou Institute of Technology, Xuzhou,Jiangsu 221008,China； 2.School of Electronic and Information Engineering ,Xuzhou Institute of Technology, Xuzhou,Jiangsu 221008,China；） Abstract: To reduce the calculation complexity of the JPDA joint-association event,which was due to multiple targets’ tracks aggregation, an improved JPDA algorithm,clustering by meanshift algorithm and optimizing confirmation matrix by Bhattacharya coefficients,was proposed.The clustering center was created by meanshift algorithm.Then the tracking gate was obtained by calculating Mahalanobis distance between the clustering center and targets’ prediction observation.The Bhattacharya likelihood matrix was created,which optimize the confirmation matrix,and the influence of association events with low probability on the JPDA joint-association event was eliminated.The results show that the new method is superior to the conventional JPDA both in computational complexity and precision of estimation in multiple targets’ tracks aggregation. Key words: multi-target tracks aggregation；Meanshift clustering；tracking gate；Bhattacharya likelihood matrix；JPDA 1 引言 多目标跟踪问题（MTT）[1]，即是量测与源目标正确关联情况下的目标状态估计问题。数据关联（DA）的实质是对量测数据的分类，即将由同一目标或杂波产生的量测归为一类。而煤矿跟踪场景由于具有背景复杂以及目标密集分布等特殊性[2,3]，目标航迹聚集情况下，量测来源存在不确定性（量测可能源于杂波或者其他目标），因此量测与源目标对应关系需要重新定义。因此有效的DA成为决定煤矿跟踪视场下MTT性能的关键因素。经典JPDA（JPDA）