计算机应用 ›› 2010, Vol. 30 ›› Issue (12): 3307-3310.

• 虚拟现实与模式识别 • 上一篇    下一篇

基于多相机的人脸姿态识别

王磊1,胡超2,吴捷2,贺庆2,刘伟2   

  1. 1. 山东大学 控制科学与工程学院;中国科学院 深圳先进技术研究院
    2.
  • 收稿日期:2010-05-18 修回日期:2010-07-08 发布日期:2010-12-22 出版日期:2010-12-01
  • 通讯作者: 王磊

Multi-camera face gesture recognition

  • Received:2010-05-18 Revised:2010-07-08 Online:2010-12-22 Published:2010-12-01
  • Contact: Lei Wang

摘要: 主动形状模型(ASM)算法被用来进行人脸特征点的精确定位,然后在多相机测量的图像中进行特征点的立体匹配,利用双目视觉和相机三维测距技术可以确定人脸特征点的空间三维位置,从而利用这些特征点的相对位置确定出人脸的姿态。实验结果显示,用该方法进行人脸姿态识别能取得比二维识别更高的精确度。

关键词: 主动形状模型, 立体匹配, 双目视觉, 特征匹配, 人脸姿态识别

Abstract: Active Shape Model (ASM) algorithm was used for the precise location of facial feature points, and then the feature points in the multi-camera images were matched to determine the spatial three-dimensional position with the technology of binocular vision and three-dimensional camera distance technology, and then human face gesture was estimated. The experimental results show that the results of the approach for face gesture recognition are very accurate.

Key words: Active Shape Model (ASM), stereo matching, binocular vision, feature matching, face gesture recognition