中文摘要：

为减少人脸表情跟踪过程中的人工干预,提高跟踪的效率和自动化程度,提出一种基于法向保持的三维人脸表情跟踪算法.采用以视频频率获取的帧人脸点云作为输入,通过将可变形的参数化网格模型依次对准每一帧点云来跟踪表情的变化;用法向保持条件自动约束跟踪过程中网格顶点在相邻帧间的运动.与已有算法相比,该算法无需在每一帧对准中人工指定特征对应点,也无需进行复杂的光流运算,可实现高效的自动化表情跟踪.另外,当参数化网格与点云所表示的人脸形状相差较大时,采用文中算法也可以得到理想的对准效果.

英文摘要：

To reduce manual intervention and improve the efficiency in facial expression tracking,a new method for 3D facial expression tracking based on normal-preserving is presented in this paper.The new method takes the time-varying face point clouds obtained at video-rate as input,and uses a deformable parametric mesh model to register each point cloud frame so as to track the varying expression.The normal-preserving condition aims to automatically constraint the intra-frame vertex motion in the tracking process.Compared with those existing facial expression tracking methods,the new method does not need manual selection of feature correspondences in each registration,and also does not need the complicated optical flow computation,so it can achieve automatic facial expression tracking with high efficiency.Additionally,when there is large difference between the shape of the parametric mesh and the point cloud,the new method could also get suggested registration result.