计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (2): 24-27.DOI: 10.3778/j.issn.1002-8331.2009.02.007

• 博士论坛 • 上一篇    下一篇

Mean-shift跟踪中的状态分析及处理

覃 剑,曾孝平,曾 浩   

  1. 重庆大学 通信工程学院,重庆 400030
  • 收稿日期:2008-09-04 修回日期:2008-10-21 出版日期:2009-01-11 发布日期:2009-01-11
  • 通讯作者: 覃 剑

Status analysis and treatment of mean-shift tracking

QIN Jian,ZENG Xiao-ping,ZENG Hao   

  1. College of Communication Engineering,Chongqing University,Chongqing 400030,China
  • Received:2008-09-04 Revised:2008-10-21 Online:2009-01-11 Published:2009-01-11
  • Contact: QIN Jian

摘要: 在Mean-shift算法中,对跟踪状态的误判会引起错误的模板调整策略,造成目标丢失。提出了一种新的跟踪状态分析方法,该算法首先通过分析目标特征与背景特征的相对关系,引入特征增强函数,并在此基础上构造了新的背景模板。然后通过对跟踪中各模板相似系数动态变化过程的综合分析,准确判定目标所处跟踪状态,以采取相应的模板调整策略。实验表明,该算法可以有效增强Mean-shift算法在复杂条件下的跟踪效果,具有较好的稳健性。

关键词: 目标跟踪, 均值漂移, 跟踪状态分析

Abstract: Misjudgment on tracking status will result in the wrong template update strategy in mean-shift and loss the object.This article presents a novel tracking status analysis method.Firstly,through the comparison of features in target and background,a feature-enhancement function is introduced and a new background template is constructed based on it.Then,after comprehensive analysis of the dynamic changing process of each similarity coefficient during tracking,the proposed algorithm can accurately determines the tracking status and takes corresponding template update strategy.Experimental results show that the proposed algorithm can effectively enhance the tracking effect under complex condition and is more robust.

Key words: object tracking, mean-shift, tracking status analysis