一种MEAN SHIFT跟踪改进算法研究  

An improved tracking algorithm based on mean shift

作　　者：李超[1] 贺鹏[1] 刘利平[1] 薛彦兵[1] 

机构地区：[1]天津理工大学计算机与通信工程学院,天津300191

出　　处：《河北省科学院学报》2012年第2期8-10,33,共4页Journal of The Hebei Academy of Sciences

基　　金：天津市科技支撑计划重点项目(10ZCKFGX00400);中国大学生创新性实验(101006020;TJUT)

摘　　要：Mean Shift算法在视频序列中的目标跟踪已经广泛被应用于计算机视觉研究以及应用中。该算法应用于跟踪中,具有计算量低,可实时跟踪等优点,但有时会出现漂移问题。本文针对传统Mean Shift算法的实现进行了研究和总结。并提出了基于分块的改进方法,在Mean Shift跟踪算法中加入了空间信息。实验证明改进算法与原始算法比较具有更好的跟踪精度。Object tracking in video stream using Mean Shift Algorithm has been widely employed in Computer Vision Research and Application.This algorithm has the feature of low computation complexity and the ability of real-time tracking,while it also has the problem of shift.In this paper,classical Mean Shift Algorithm has been researched and summarized,and a new method that partition the big block into small pieces is asserted,which add spatial information into the Tracking Algorithm.Some experimental result prove that the improved algorithm is more accurate than the classical one.

关 键 词：跟踪 Mean SHIFT 分块 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]