摘要
针对单张人像的三维姿态计算,结合面貌测量和射影几何的理论提出了一种方法:首先在人面部的平面区域内,选取眼角点,口角点,鼻翼点建立人脸模型;然后根据人脸平面上两个相互垂直的特征线投影到照片上的灭点位置,求出人脸平面的旋转方向。该方法特征点易于标定,且无需任何的辅助设备和先验知识,具有一定的实用性。
Aiming at 3D pose computation from single facial image, a method was proposed based on the theory of anthropometry in face and projective geometry first, comers of eyes and mouth, and nosewing points were selected to establish facial model within facial plane; then, the orientation of face was determined based on the leoafion of varfishing points in image, which were resulted from two perpendicular features lines in perspective project. This method can be implemented without any auxiliary equipments and prior knowledge, distinct feature points are easy to be located, and the practicability can be verified.
出处
《计算机应用》
CSCD
北大核心
2007年第5期1138-1140,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(60573179)
关键词
姿态估计
颅像重合
灭点
人脸
pose estimation
photographic superimposition
vanishing points
