The Study Of 3d Face Recognition System

In recent decades, biometric technology is widely applied to various fields. Face recognition is the most active and most challenging task based on biometric authentication technology, also one of the most potential technologies in the century. The 3d face recognition is expected to solve 2d face recognition bottlenecks of effecting identifying by lighting, poses and expressions .Therefore it becomes the focus of many institutions research. However, the current domestic and foreign institutions research focus on the practical application of the system are not very sufficient. Therefore, the paper will study the complete 3d face recognition system.To achieve research request, this article made the following job:1 First of all, this paper introduced the implications of the research of face recognition system and the present research of main methods and key technologies. The difficulties of the topic research are provided ideas for further research.2 The traditional ASM animation - face feature points algorithm is too time-consuming and accuracy is not high. The traditional algorithm is improved and optimized. In the traditional algorithm first matching stage .Put forward based on the face triangle feature point positioning method. In addition to the algorithm search process, joined the face of global texture information constraint, which made the algorithm time-consuming greatly decreased. Improved face feature point positioning more accuracy. For 3d face model of laying the foundation.3 According to the binocular stereovision theory, created a preliminary 3d face models. At first, completed camera calibration, then through the two tickets to each Angle face 2d images on different coordinate 2d coordinate, synthesized 3d coordinates under 3d coordinates, established facial 3d model of contour.4 The traditional binocular stereovision establish 3d face models in extracting feature points, 3d feature point is limited, can not restore face 3d features point cloud information. In order to solve this problem, this paper proposed an automatic iterative interpolation algorithm. The algorithm was able to automatic iterative calculation relatively accurate 3d face data points. As far as possible to recovery 3d face feature point cloud data .Established saturation higher 3d face models which can be used for identification. 5 In view of the current based on the face of the 3d geometric characteristics identification algorithm, which cannot be very good solve identification accuracy when face expression changed. This paper presented a based on geometric characteristic changes association similarity degree of 3d face recognition method. Use relevant knowledge of Probability learn, computational geometry characteristic difference degree. Through calculating contains expression changes association similarity degree of similarity to 3d face recognition, improved the 3d face recognition accuracy.Experiments proved that this 3d face recognition system can efficiently identified 3d face.