New approach on PCA-based 3D face recognition and authentication

In this paper, we propose a new approach which allows us producing a new representation independent from the position and the orientation of each 3D point cloud. This approach builds, from the 3D point cloud, the models of faces which are used afterward for the recognition. This framework allows us to use statistical inferences such as the estimation of the missing parts of the face by means of the PCA on the tangent spaces of the variety of shape. For that purpose, the proposed method explores the basic of projection by comparing every point cloud input with those of the database. To reduce the cost of the exploration, we define a comparison function based on the transformed of 3D distance. Experimental results using real data show the potential of our method, we obtain a 99% rate of verification performance of the CASIA-3D dataset, which compares well with other state of the art methods.