Research On Conformal Mapping Algorithm-based3D Face Recognition

Using computers automatically to recognize human face has a wide rangeof uses in human society. While humans can easily depend on facial featureto identify some person’s identity, computer automatically recognizingface is a very difficult problem. At present, the main methods of facerecognition are based on2D images, however changes in illumination,posture, facial expressions and other factors will significantly reducethe performance of2D face recognition algorithm. Therefore, people areturning their attention from2D face recognition to3D face recognition.With the development and popularization of3D data acquisition device,people easier acquire a high-quality3D face model.3D face model is notsensitive to the change in illumination, and has3D geometric featuresof depth, curvature, etc.3D face recognition also has the potential toovercome the difficulties of2D face recognition caused by change in faceposture. Consequently,3D face recognition is considered to be aneffective way for face recognition.In this paper, the problem of3D face recognition based on harmonicmapping algorithm has been studied. After face region extraction, surfacesmoothing and hole repairing, the original3D face model becomes a neat3D face mesh. We convert it into a2D face image by harmonic mapping.Finally the2D face images are matched and recognized by a2D imagerecognition method. The main work of this paper are as follows:3D Face Segmentation and Extraction: A new method of crude extractionof3D face region which is light-based has been proposed. We find out the shadow area of the3D face under two light sources, and then detect thebinding site between the main area of face and the area of shoulder withneck. After all, we can cut up the above two parts and extract the mainarea of face.3D Face Surface Smoothing: A grid subdivision algorithm based on theLoop subdivision algorithm has been implemented. The grid subdivisionalgorithm can achieve the goal of surface smoothing and make theaccidented3D face model becomes a new one with continuous and stablesurface shape.3D Face Holes Repairing: The repairing face holes algorithm based onthe wave-front method has been implemented. Before holes repairing, usingthe grid subdivision method to process the accidented3D face can enhancethe repairing effect of wave-front method.3D Face Posture Correction and Face Reconstruction: The posture offaces in non-frontal view have been corrected, and using the symmetricalcharacteristics of face, the missing data of faces in non-frontal viewhas been filled up.3D Face Recognition: A3D face recognition algorithm based on harmonicmap algorithm and (2)2has been implemented. The recognitionexperiment has been carried out. The experimental result demonstrates theeffectiveness of such3D face recognition algorithm.