基于分块与前景/背景信息的MeanShift目标跟踪

针对经典Mean Shift跟踪算法在目标遮挡和复杂背景情况下易造成目标漂移甚至丢失的缺点,该文研究了基于分块与前景/背景信息的Mean Shift跟踪方法.首先根据实际目标尺寸对跟踪窗口进行分块,然后对每个子块独立进行Mean Shift跟踪,最后按照一定的准则融合每个子块的跟踪结果以确定整体目标的位置.并且,通过目...

Robust Object Tracking Using Fragments-based Mean Shift and Foreground/Background Information

A robust tracking algorithm is proposed to solve the problem that traditional mean-shift based trackers always drift or even lose the targets under occlusions and clutter.The target is divided into several fragments based on target＇s shape and size,the mean shift algorithm is used for every fragment to find a best matching location independently.The confidence of each fragment is computed and only the fragments with higher confidence are involved to determine the location of the entire target.Furthermore,weights give to different colors in tracking window to separate reliable parts of object from its background based on foreground/background histogram analysis.The proposed algorithm is test on several challenging videos,and encouraging results indicate the efficiency of the algorithm in both handling occlusion and clutter.