The Research On Feature Extraction And 3D Reconstruction Of Human Facial For Near-Infrared Images

In recent years, computer-aided human face processing has become a hot issue in fields like computer vision, computer graphics, pattern recognition, and more. This technology can be widely applied to movie production, video conference, intelligent human computer interface, etc.In this dissertation, we discussed in detail the design and implementation of a facial feature extraction and 3D reconstruction system based on the near-infrared camera. Generally, the above-mentioned system should be composed of the following modules: camera calibration, feature detection and extraction, 3D reconstruction and 3D display. Our study focused mainly on two aspects, namely, the key technology in the feature detection and extraction module and the overall framework and implementation of the prototype system.Firstly, this dissertation summarized the current research status and existing problems while the structure and the main contents of the dissertation were briefly introduced. In the second chapter, we gave detailed overviews on several mainstream methods in facial feature extraction and 3D reconstruction. The design and implication of the entire system was discussed in the third chapter. According to the face imaging features of the near-infrared camera, we put forward two methods which might improve the experimental results of the feature extraction module. The two methods, which were based on the geometry characteristics of the marker and the constraint of Epipolar Geometry respectively, were described in great detail. Besides, we elaborated how Intel's Open Computer Vision Library (OpenCV) was adopted to implement the system. At last, this dissertation analysed the results of the experiment, discussed the possible values of the three key factors which significantly affected the experimental results, and gave the final conclusions.