中国科学院心理研究所机构知识库

Face recognition is still one of the research hotspots in the field of pattern recognition and artificial intelligence. Compared with two-dimensional face images, three-dimensional face model can provide key depth information, which is less affected by pose and illumination, and has better adaptability to the change of expression. Through the research of 3D face recognition method, the difficulties of 2D face recognition are overcome. Aiming at the shortage of the existing 3D face data set, this paper designs a 3D face recognition method, and proposes a data set synthesized by 3dmm. At the same time, by transforming the expression and shape on the data set, enough face samples can be synthesized to obtain 3D face point cloud data with good expression ability, the accuracy of face recognition is improved by the 3D face recognition technology of multi feature fusion. For 3D face data, this paper uses face point cloud, which avoids the loss of key information when mapping from 3D face to 2D face, and avoids the huge data of 3D face voxel network.