北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2006, Vol. 29 ›› Issue (s2): 148-150.doi: 10.13190/jbupt.2006s2.148.mej

• 论文 • 上一篇    下一篇

基于光流的运动人体提取

么 键, 刘冀伟, 韩 旭, 王志良   

  1. 北京科技大学 信息工程学院, 北京 100083
  • 收稿日期:2006-09-05 修回日期:1900-01-01 出版日期:2006-11-30 发布日期:2006-11-30
  • 通讯作者: 么 键

Gait Extraction Based on Optical Flow

YAO Jian, LIU Ji-wei, HAN Xu, WANG Zhi-liang   

  1. School of Information and Engineering, Beijing University of Science and Technology, Beijing 100083, China
  • Received:2006-09-05 Revised:1900-01-01 Online:2006-11-30 Published:2006-11-30
  • Contact: YAO Jian

摘要:

提出了一种改进的Lucas-Kanade光流算法,将这种算法应用于步态识别中, 并引入了高阶梯度的光流场约束方程,计算出其Hessian矩阵. 将Hessian矩阵条件数的倒数做为Lucas-Kanade光流法的加权阵,可有效地消除局部邻域中不可靠约束点,并大大提高基本约束方程解的稳定性. 实验结果表明,此方法能精确地描述出运动人体的光流场.

关键词: 步态识别, 光流, 梯度

Abstract:

An improved technique for Lucas-Kanade optical flow was proposed, which was used in the field of gait recognition. Firstly, a constrained equation of high-order gradient was introduced, and then Hessian matrix can be computed. Secondly, the reciprocal of conditional operators of Hessian matrix was taken as the weighted matrix of Lucas-Kanade flow field. This method can eliminate the unreliable constrained points of local neighborhood, and increase the stability of the basic solution of the constrained equation. The experiment results showed that the proposed method can establish the optical flow field for human gait precisely

Key words: gait recognition, optical flow, gradient

中图分类号: