视频序列中面向人的多目标跟踪算法  

Human oriented multi-target tracking algorithm in video sequence

作　　者：常发亮[1] 马丽[1] 乔谊正[1] 

机构地区：[1]山东大学控制科学与工程学院,济南250061

出　　处：《控制与决策》2007年第4期418-422,共5页Control and Decision

基　　金：国家自然科学基金项目(60104009);山东省自然科学基金项目(Z2005G03)

摘　　要：针对视频序列中人的跟踪问题,提出一种基于运动检测的多目标跟踪算法.跟踪系统由运动目标检测、关联矩阵建立、特殊情况判断及处理以及轨迹关联4部分构成.提出一种基于改进的c-均值聚类的自适应运动分割方法;不同情况下建立不同的关联矩阵,以准确判断实际场景状况;对遮挡问题作出处理,在两个目标遮挡不严重的情况下,分别采用均值漂移算法对其进行跟踪.实验结果表明,该算法具有较强的鲁棒性,能有效实现复杂场景下多目标跟踪.Aiming at tracking human in video sequence, a multi-target tracking algorithm based on moving detection is proposed. Tracking system is composed of four parts： Moving targets detection, association matrix forming, special cases estimation and handling, and track association. An adaptive moving segmentation method based on improved c- mean clutter is presented. Different association matrix in different condition is used to estimate the actual scene states accurately. Handling for occlusion is achieved, and in the case of mild occlusion of two targets, mean shift algorithm is applied to track them separately. Experimental results show that it is robust and has good multi-target tracking performances under complex background.

关 键 词：多目标跟踪 运动检测 轨迹关联 遮挡 均值漂移 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]