人脸三维测量中的立体视觉

Stereo Vision in 3D Facial Morphometry

探讨了立体视觉在人脸外形测量中的应用，针对牙齿整容手术模拟时必须测量面部皮肤表面一些指定点三维信息的问题，建立了一种双目成象系统．根据人脸的特征，提出了人机交互确定特征点并结合先验知识进行图象匹配的新方法，通过受试者的正面和左右９０°侧面三组图象、成象系统的全局坐标系与面部局部坐标系之间的几何位置，确定指定点在面部局部坐标系中的坐标值．误差校正项的引入有效地提高了系统的精度，实验结果表明了该系统的合理性．

The application of stereo vision in 3D facial morphometry is investigated. A binocular image system is presented for measuring 3D values of some marked cutaneous points on the human face in dental surgery simulation. According to the facial feature, an interactive method in recognition of marked points is adopted. Stereo matching is made by prior knowledge.By the images acquired from the front, 90° right side and left side of a subject, and by the geometric relation between global coordinate system of the model and the local coordinate system on the subject face, the positions of facial landmarks are collected in the local coordinate system. The error correction improves the accuracy of the facial morphometry. The experiment results exhibit the effectiveness of the system.