Research On 3D Face Recognition Based On Gabor Wavelet And Support Vector Machine

Human face is intrinsically 3D deformable object with texture. The 3D shape information should not be ignored in face recognition since they provide another type of distinct feature to distinguish different faces. The research points in 3D face recognition mainly include 3D data acquisition,pretreatment,feature extraction and classifier design.In this thesis, based on the depth image, we focus on the following problems: feature extraction and classifier design, and the method of 3D face recognition based on Gabor wavelet transform and Support Vector Machine is proposed. 3D face contour feature extracted by using multi-channels Gabor wavelet filters was denoted by the coefficients of Gabor wavelet transform and its standard variance. In addition, the feature vectors are used to train and identify by support vector machine. Here lists the main work and our contributions in this thesis:1,As 3D point-cloud data are scattered points without structural information, the grid control points are used to simulate the point-cloud data which is created by B-spline surface fitting. We standardize the point-cloud data to reduce the quantity of point-cloud data to raise efficiency and robustness of our following algorithm.2,In the aspect of feature extraction, a method of 1D Gabor wavelet feature extraction based on 3D face contour is proposed. First, face contours are extracted in the basis of 3D face coordination system with the depth information, then viewing the contours as 1D signals, do the work of 1D Gabor wavelet to own great discriminating power for 3D face feature.3,We propose a simple but effective method, Support Vector Machine, to discriminate 3D faces. The experiment results show that the algorithm of 1D Gabor wavelet based on 3D face contours and SVM classifier, has a fast calculating speed, and the face recognition rates are around 86.7% on better face images and 81.5% on the whole 3D face database of ZJU-3DFED.