作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2011, Vol. 37 ›› Issue (14): 158-160. doi: 10.3969/j.issn.1000-3428.2011.14.052

• 人工智能及识别技术 • 上一篇    下一篇

基于UKF的窗口自适应Mean-Shift算法

杨 帆,郑春红,杨 刚   

  1. (西安电子科技大学电子工程学院,西安 710071)
  • 收稿日期:2011-01-03 出版日期:2011-07-20 发布日期:2011-07-20
  • 作者简介:杨 帆(1986-),男,硕士研究生,主研方向:图像处理,目标跟踪,目标识别;郑春红,副教授;杨 刚,教授

Mean-Shift Algorithm with Adaptive Window Based on UKF

YANG Fan, ZHENG Chun-hong, YANG Gang   

  1. (School of Electronic Engineering, Xidian University, Xi’an 710071, China)
  • Received:2011-01-03 Online:2011-07-20 Published:2011-07-20

摘要: 传统的Mean-Shift跟踪算法窗口固定,不能对尺度任意变化的目标进行有效跟踪。为此,提出一种多尺度理论与无味卡尔曼滤波器(UKF)相结合的视频跟踪改进算法。利用多尺度理论统计跟踪窗内的信息量,使用UKF对得到的信息量进行预测,通过修正后的信息量计算窗口变化比例系数,对尺度任意变化的目标进行跟踪。实验结果证明,该算法能对尺度任意变换的目标进行有效跟踪。

关键词: 目标跟踪, 无味卡尔曼滤波器, Mean-Shift算法, 信息度量

Abstract: The traditional fixed bandwidth Mean-Shift tracking algorithm can not have an effective tracking for any changes in targets. An novel method is proposed that is multi-scale space theory combined with Unscented Kalman Filter(UKF). UKF filter is introduced to predict the information in the tracking window which is calculated by the multi-scale space theory. The proportion of the target image area is got by the modified information. It is implemented by the combination of the Mean-Shift tracking algorithm and UKF to track targets. Experimental result confirms the effectiveness of the improved algorithm.

Key words: object tracking, Unscented Kalman Filter(UKF), Mean-Shift algorithm, information measure

中图分类号: