The Study Of 3D Face Recognition Based On Facial Surface Polar Spectrum

Face recognition got widely concerned in last twenty years. Many methods have been proposed. However, most of them are based on 2D image, and are affected by the luminance, face pose and gesture, which limit the performance of application.With the improvement of 3D acquisition technology, more and more researchers proposed recognition methods based on 3D face model. 3D face model represents more clues for recognition than 2D image, and can be tolerant to the influence of luminance and face pose. The main contributions of the work are as follows:1) We address a new description for 3D face approach based on the polar-spectrum image of flattened facial surface. Firstly, the 3D face is parameterized into an isomorphic 2D planar circle, attempting to preserve the intrinsic geometric properties. Then, the flattened surface transforms using FFT in polar-coordinate domain. The description is pose-invariant and decreases the affection of expression.2) We proposed a 3D face recognition algorithm using Fisher-Spectrum. The experiment demonstrates its comparable performance on 3DJRMA.3) Our previous works demonstrated the profile was a good characteristic for 3D face recognition. So, the experimental results using robust profile normalization method, profile similarity metric for robust profile recognition are given.