基于协同关联粒子滤波算法的交互多视频目标跟踪(英文)  

Collaborative associated particle filter for interactive multi-target tracking in video surveillance

作　　者：韩华[1,2] 丁永生[1,3] 郝矿荣[1,3] 

机构地区：[1]东华大学信息科学与技术学院,上海201620 [2]上海工程技术大学电子电气工程学院,上海201620 [3]东华大学数字化纺织服装技术教育部工程研究中心,上海201620

出　　处：《控制理论与应用》2013年第9期1187-1193,共7页Control Theory & Applications

基　　金：supported by the Key Project of the National Nature Science Foundation of China(No.61134009);the National Nature Science Foundation of China(Nos.61272097,61305014);Specialized Research Fund for Shanghai Leading Talents,Project of the Shanghai Committee of Science and Technology(No.11JC1400200);the Fundamental Research Funds for the Central Universities,Innovation Program of Shanghai Municipal Education Commission(12ZZ182);the Nature Science Foundation of Shanghai(13ZR1455200);Funding Scheme for Training Young Teachers in Shanghai Colleges(ZZGJD13006)

摘　　要：首先介绍了带马尔科夫跳变非线性系统(JMNSs)的状态估计问题,然后总结了JMNSs最优状态估计的难点和具有交互作用的多目标跟踪问题.在总结分析各类不同算法的基础上,提出了一种协同关联粒子滤波算法来解决目标数目在变化的交互多目标跟踪问题,改进后的算法不需要观测与目标状态关联和目标数量已知的假设.最后,通过仿真实验验证了改进后的算法在跟踪效果上优于现有算法,并能成功估计目标的数量.We first introduce the state estimation of jump Markovian nonlinear systems （JMNSs）, with a summary of difficulties in this estimation; and then we review the problems of the interactive multi-target tracking. Based on the analysis of various algorithms, a collaborative associated particle filter is proposed to solve the problem of interactive multi- target tracking with time-varying target numbers. The proposed algorithm neither needs the assumption of the association of observations with target states, nor the knowledge of the target numbers. Simulation results show that the proposed algorithm provides better tracking performances and more accurate estimation of the target numbers.

关 键 词：粒子滤波 协同理论 数据关联 多目标跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]