Different Expressions And Postures3D Face Recognition Based On Iso-geodesic Curves

Because of face recognition based on2D image can’t conquer the impact brought by illumination, posture, makeup, age and the expression change etc, there is still not a robust automatic identification system. Along with the point cloud data acquisition becomes more and more easily,3D face recognition has become a hot research topic in recent years. The change of expression and posture is the most difficulty of3D face recognition. In recent years, research suggests that expression variation can be modeled as similar isometric transformations. The traditional expression invariant based on geodesics distance is obtained by discrete sampling on iso-geodesic curves, with only local features, lacking of global features. In this paper, we introduce five shape descriptors on the base of the original isometric features to overcome the limitation by offering global geometric features. Furthermore, the shape descriptors are translation and in-plane rotation invariant, for aligned2.5depth image, they are also pose invariant.First, using the method and process proposed in this paper, we make different postures of3D point data have well registrations. Then, we get depth and intensity images from the aligned3D data. Next, we compute the radial geodesic distance from all the points to the nose tip in depth image, thus, distance image and level curves can be generated. On the one hand, we construct texture invariant feature vectors from intensity values in intensity image at fixed points; on the other hand, we get the level curve’s shape descriptors from the depth images. They form this article’s invariant feature vectors together. At last, recognition work is completed using invariant feature vectors by LDA and Mahalanobis distance. Our experiment result shows that our algorithm performs better than the traditional method based on geodesic distance and the advanced method based on diffusion distance.