基于Candide-3模型的姿态表情人脸识别研究

针对姿态表情严重影响人脸识别准确率的问题,基于Candide-3模型的简化,提出了形状表情关键点拟合的人脸几何结构重建和基于三角网格模型的纹理映射的方法,该方法确定关键特征点,根据人脸的几何结构信息确定姿态角,提取Candide-3模型形状表情对应点,调整模型参数,进行几何结构重建；对几何结构中每个三角网格模型进行纹理影射,得到逼真的特定人脸模型.实验结果表明,该方法提高了人脸重建速度,达到减弱姿态表情对人脸识别影响的目的.

Study of pose and expression variant face recognition based on Candide-3 model

For pose and expressions seriously affecting to the accuracy of face recognition,based on the Candide-3 model,a method of pose and expression key points fitting to reconstruct face geometry and texture mapping based on Triangular mesh model are proposed.First the eyes,nose,mouth feature points are determined according to the face geometry information,then the pose angle are determined,facial expressions corresponding points are extracted with Candide-3 model,the model parameters are adjusted to make structural reconstruction,then texture mapping of each triangle mesh in the geometry are completed to obtain specific 3D face reconstruction model.Experiment shows that the method improve the speed of 3D face reconstruction and the effect of pose and expression variant to face recognition is reduced.