Estimation Of Face Rotation And 3D Facial Recognition Based On Depth Data

As a kind of biometric technology, face recognition is one of the most active and potential research areas in the fields of pattern recognition and computer vision and it has broad application prospects in commerce, security and other fields. 3D face recognition based on 3D data, combined with computer vision and computer graphics, takes full advantage of depth information of 3D face. Therefore, 3D face recognition is able to overcome the problems resulting from illumination, expression and pose variations in 2D face recognition. The estimation of face rotation is the preprocessing step of 3D face recognition and the prerequisite for intelligent human-computer interaction and face recognition. This thesis is mainly focused on the estimation of face rotation and face recognition based on 3D face depth information. Main work of innovation is listed as follows:1. The overview of the conception and basic procedure of 3D face recognition is given, the 3D face data acquisition and representation and the evaluation of face recognition systems are also concerned. The algorithms of 3D face recognition based on depth information are introduced in details.2. A novel method of calculating the rotation of face is proposed based on the 3D facial depth information. Facial feature points are extracted using the depth map and its corresponding gray image according to differential geometry theory, curvature algorithm. Using intensity features generated by face data, right and left pupil can be located. And then the three pose angles of the face in the 3D space are calculated.3. Given face rotation angles, a novel method of 3D face recognition based on profiles is proposed. Firstly, the depth map after posture correction can be calculated by rotating the original face posture utilizing principle of computer graphics to the frontal state and regularizing the corrected facial depth data after surface fitting and re-sampling. Then two important surface contours are extracted: center profile and the cross section contour at nose tip. Finally, the result of recognition is obtained by using iterative closest points (ICP) algorithm to register the profiles.