计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (27): 50-53.

• 学术探讨 • 上一篇    下一篇

跟踪遮挡目标的一种鲁棒算法

王展青1,2,凡友福2,张桂林1   

  1. 1.华中科技大学 图像信息处理与智能控制教育部重点实验室,武汉 430074
    2.武汉理工大学 理学院,武汉 430070
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-09-21 发布日期:2007-09-21
  • 通讯作者: 王展青

Robust algorithm for occlusion problem in tracking

WANG Zhan-qing1,2,FAN You-fu2,ZHANG Gui-lin1   

  1. 1.State Education Commission Key Laboratory for Image Processing and Intelligent Control,Huazhong University of Science and
    Technology,Wuhan 430074,China
    2.College of Science,Wuhan University of Technology,Wuhan 430070,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-09-21 Published:2007-09-21
  • Contact: WANG Zhan-qing

摘要: 为了解决在跟踪目标过程中的遮挡问题,引入Kalman滤波器为Mean Shift跟踪算法选择初始点,在跟踪稳定的情况下进行模型更新以消除由于目标缓慢变化而产生的累积误差对跟踪结果的影响。根据Kalman滤波器残差的大小判定是否发生遮挡,遮拦检测算法对目标进行分块检测从而把遮挡分为部分遮挡和完全遮挡两种情况,并对两种情况进行区别讨论:对部分遮挡情况不做特殊处理;对完全遮挡情况,结合目标的运动方向提出6点搜索策略来找回目标。实验表明,该算法能很好地解决跟踪运动目标过程中目标的遮挡问题。

关键词: Mean Shift, 卡尔曼滤波器, 目标跟踪, 遮挡检测

Abstract: We employ Kalman filter to choose the start point for Mean Shift tracking algorithm and update the target model to avoid the accumulative error result from the tardy change of target.Kalman filter error is exploited to decide whether the occlusion detection algorithm works,which divides the target into four parts and detects respectively.And occlusion is classified into two categories: partial occlusion and complete occlusion,we utilize the velocity of the target to present six-point research strategy to deal with complete occlusion.Experimental results show that the algorithm is robust to solve occlusion problem.

Key words: Mean Shift, Kalman filter, object tracking, occlusion detection