改进的Meanshift运动目标跟踪算法  

Improved meanshift moving tracking algorithm

作　　者：张英[1] 车进[1] 牟晓凯[1] 白雪冰[1] 

机构地区：[1]宁夏大学物理电气信息学院,宁夏银川750021

出　　处：《电视技术》2016年第10期97-100,126,共5页Video Engineering

基　　金：国家自然科学基金项目(61162020)

摘　　要：Meanshift算法在对快速运动的目标进行跟踪时容易丢失目标,并且在目标被遮挡时,也容易造成跟踪失败,跟踪的过程中跟踪框不能随着运动目标的大小变化而变化。提出一种基于Meanshift运动目标跟踪算法的改进算法。该算法基本思想是采用改进的三帧差分法对运动目标区域进行提取,求得跟踪框轮廓,同时用Meanshift算法对运动目标进行跟踪,获得目标最大概率区域,将该区域中心作为跟踪框的中心。跟踪过程中通过巴氏系数判断是否目标被遮挡,若被遮挡则调用Kalman滤波进行预测跟踪。实验结果表明,该算法能够快速、准确地跟踪目标。When Meanshift algorithm tracks the fast-moving targets, it is easy to lose the target. And when the targets are blocked, it is likely to cause tracking failure. The tracking window size can not be adapted to change. An improved Meanshift target tracking algorithm is presented. The algorithm uses improved three-frame difference method to extract moving target area, and seek tracking frame outline. At the same time, Meanshift algorithm tracks the moving targets and obtains the target maximum probability region. Then the regional center can be a center of the tracking frame. During tracking process, Pap coefficient can determines whether the target is blocked. If target is blocked, then kalman filter can be called to forecast and track. Experimental results show that the al- gorithm is quick and accurate to track the target.

关 键 词：三帧差分 MEANSHIFT 运动目标跟踪 KALMAN滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]