Computer Engineering and Applications ›› 2010, Vol. 46 ›› Issue (16): 200-203.DOI: 10.3778/j.issn.1002-8331.2010.16.058

• 图形、图像、模式识别 • Previous Articles     Next Articles

Mean Shift tracking algorithm with hybrid target model

LI Yun-han,LOU Jing-jing   

  1. Department of Computer Science and Engineering,Yiwu Industrial & Commercial College,Yiwu,Zhejiang 322000,China
  • Received:2009-02-09 Revised:2009-04-20 Online:2010-06-01 Published:2010-06-01
  • Contact: LI Yun-han

混合目标模型的Mean Shift跟踪算法


  1. 义乌工商职业技术学院 计算机工程系,浙江 义乌 322000
  • 通讯作者: 黎云汉

Abstract: As Mean Shift algorithms with unchanged target model matching candidate by histogram,the histogram is a relative weak characteristic,the accurateness of tracking is unsatisfactory when the background and the target have similar color attribute.To improve the limitation of Mean Shift,a modified algorithm with hybrid target model is proposed.The hybrid target model takes into account the information of the first frame and the previous frame,thus overcomes the weakness of unchanging target model that can not accurately represent the rotate target and the target that has similar color with background.Experimental results show that the proposed algorithm can achieve better effect.

Key words: hybrid target model, Mean Shift, object tracking

摘要: 但固定目标模型的Mean Shift算法采用直方图进行匹配,而直方图是一种比较弱的目标特征,当背景和目标的颜色分布较相似时其跟踪效果欠佳。针对这一缺点,提出了一种采用混合目标模型的Mean Shift算法。该算法在匹配过程中使用的目标模型包含了初始帧和前一帧的信息,克服了固定目标模型难以对与背景相似目标以及旋转目标进行准确描述的缺点,获得了较好的跟踪效果。

关键词: 混合目标模型, Mean Shift, 目标跟踪

CLC Number: