3D Face Recognition Base On Multimodal Fusion

3D face recognition using the data come from the3D point cloud, which is less sensitive to expressions and pos changes.On the other hand, it means we discarded the texture image information from the face. In this paper here, we present a algorithm fusion2D and3D face information, and have robust to expressions and pos changes. First, we use ICP normalize the3D face point cloud data, and the we project the point set on to2D image, and we can have2D face image with proper alignment。We extract3D features base on statistics information, using ICP normalize the features extracted from different face, the compute the Hausdorff distance between them. In2D case we compute the SIFT base feature from different image. And last we fusion the two classifiers in the score level, the proposed algorithm achieved94.1%identification rate on the part of CASIA3D Face Database (240image).