基于目标中心定位和NMI特征的跟踪算法  

AN TRACKING ALGORITHM BASED ON OBJECT CENTRE ALLOCATION AND NMI FEATURE

作　　者：陈爱斌[1,2] 董德毅[1] 杨勇[1] 蔡自兴[2] 

机构地区：[1]中南林业科技大学计算机科学学院,湖南长沙410004 [2]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《计算机应用与软件》2010年第4期276-279,共4页Computer Applications and Software

摘　　要：实时目标跟踪过程中,为提高跟踪精度,要求跟踪的窗口也要实时的随着目标大小的变化而变化。以Mean shift算法为基础,根据概率检测法定位目标在各帧图像中的中心点,提出了一种自适应更新窗口的算法。同时结合归一化转动惯量NMI对目标进行识别,实时定位中心标示的对象。实验结果表明,该方法能在目标尺寸放大或缩小时选择合适的跟踪窗口,而且具有较强的跟踪抗干扰性。In order to improve the tracking precision during the process of real-time target tracking,the tracking bandwidth is required to vary timely along with the size altering of the objects.Based on the algorithm of mean shift and on locating objects’central points in each frame of image according to probability detection method,the thesis puts forward a self-adaptive algorithm of bandwidth updating.Also the normalized moment of inertia（NMI） feature is combined for identifying the object to locate in real-time the object designated by the centre.Experiments show that this method can choose proper size of tracking bandwidth when the objects＇size are altering and has strong anti-jamming property.

关 键 词：目标跟踪 均值偏移 目标中心 NMI特征 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]