计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (13): 194-196.DOI: 10.3778/j.issn.1002-8331.2009.13.057

• 图形、图像、模式识别 • 上一篇    下一篇

基于背景提取和扩展均值漂移算法的目标跟踪

曹玉华1,吴小俊2,段先华2,王士同2   

  1. 1.江苏科技大学 电子信息学院,江苏 镇江 212003
    2.江南大学 信息工程系,江苏 无锡 214122
  • 收稿日期:2008-03-12 修回日期:2008-05-30 出版日期:2009-05-01 发布日期:2009-05-01
  • 通讯作者: 曹玉华

Background subtraction and extend mean shift algorithm for object tracking

CAO Yu-hua1,WU Xiao-jun2,DUAN Xian-hua2,WANG Shi-tong2   

  1. 1.School of Electronics and Information,Jiangsu University of Science and Technology,Zhenjiang,Jiangsu 212003,China
    2.School of Information Engineering,Southern Yangtze University,Wuxi,Jiangsu 214122,China
  • Received:2008-03-12 Revised:2008-05-30 Online:2009-05-01 Published:2009-05-01
  • Contact: CAO Yu-hua

摘要: 通过在静态背景模型下利用自适应背景提取和扩展均值漂移算法相结合的方法对人机交互式的目标跟踪作了进一步的改进。首先利用自适应的背景提取算法从带有运动目标的复杂背景中构建背景图,并提取出运动目标轮廓。在跟踪模块,在均值漂移算法的基础上加入协方差得到的扩展均值漂移可以很好地解决传统均值漂移算法在跟踪过程中因为目标的形状或大小改变而导致跟踪的框架偏离目标的问题。实验结果表明,该算法能够较好地实现自动、实时、较准确的跟踪目标效果。

Abstract: Human computer interactive target tracking is improved by combining adaptive background extraction in static backg-round with extend mean shift algorithm.Firstly build a background image in the complex background with a moving object and obtain initial object contour from the current image by frame-difference.When tracking by mean shift algorithm,the covariance matrix is added to mean shift,and it can solve the problem of the tracked object departure.The experiments indicate that the algorithm can track the object accurately and effectively.