作者: Qingxiang Wang , Yanhong Yu
关键词:
摘要: AbstractFacial features tracking is widely used in face recognition, gesture, expression analysis, etc. AAM Active Appearance Model one of the powerful methods for objects feature localization. Nevertheless, still suffers from a few drawbacks, such as view angle change problem. We present method to solve it by using depth data acquired Kinect. use get head pose information and RGB match result. establish an approximate facial 3D gird model then initialize subsequent frames with this information. To avoid local extremum, we divide into several parts poses closest model. The experimental results show improvement performance when rotating head.