计算机应用 ›› 2011, Vol. 31 ›› Issue (01): 254-257.

• 模式识别 • 上一篇    下一篇

窗宽自适应Mean-Shift跟踪算法

白向峰1,李艾华2,李喜来1,蔡艳平1   

  1. 1. 第二炮兵工程学院
    2. 西安第二炮兵工程学院502教研室
  • 收稿日期:2010-07-13 修回日期:2010-08-25 发布日期:2011-01-12 出版日期:2011-01-01
  • 通讯作者: 白向峰

Mean-Shift tracking algorithm based on adaptive bandwidth

2, 2, 2   

  • Received:2010-07-13 Revised:2010-08-25 Online:2011-01-12 Published:2011-01-01

摘要: 针对固定窗宽Mean-Shift算法在目标运动速度过快或尺度发生明显变化时可能导致跟踪失败的问题,提出一种窗宽自适应的Mean-Shift跟踪算法。该方法基于均值漂移矢量预测跟踪窗口中心位置,同时自动调整跟踪窗口大小,保证目标始终处于跟踪窗口内部,使算法得以准确定位目标;在确定空间位置后,利用基于Bhattacharyya系数的二分法自动选取窗口缩放比例,得到与目标尺度一致的跟踪窗口。实验结果证明,该方法能很好地定位目标的空间位置和尺度。

关键词: 目标跟踪, 均值漂移, 窗宽自适应, Bhattacharyya系数

Abstract: Mean-Shift algorithm with fixed bandwidth often fails in tracking an object that moves too fast or owns a dramatic change in scale. To solve the problem, a novel Mean-Shift tracking algorithm based on adaptive bandwidth is proposed. Mean-Shift vector is used to predict the center position and automatically modulate the size of tracking window that fix the object inside the window and gain accurate object position. After the confirming of position, a Bhattacharyya coefficient based dichotomy is adopted to select the pantograph ratio automatically, and a tracking window adapt to the scale of object is obtained. Experiment results proved the algorithm’s capability in locating object’s position and scale.

Key words: object tracking, Mean-Shift, adaptive bandwidth, Bhattacharyya coefficient