The Research Of3D Face Recognition Technology Based On Depth Image From Kinect

With the rapid development of pattern recognition technology, biometric identification technology now is gaining attention from more and more researchers. The existing biometric technology mainly include recognition of hand, face recognition, fingerprint identification, voice recognition, and gait recognition, retina identification, iris recognition, etc. Due to face contains the rich biometric information, Face recognition technology has become an important biological characteristics identification technology for identity authentication and one of the most potential research directions of pattern recognition and machine vision areas. Face recognition technology has a very broad application prospects in safety test system, intelligent human-machine interface, video conference, etc.Now face recognition technology research mainly focus on the2D images, which is extremely susceptible to natural light, posture, facial expression change, make the recognition accuracy is very limited.3D model holds more rich information than2D image, so implementing face recognition on3D face model is one of the effective app. The mainly key problems needing to be solved in3D face recognition technology research includes3D face data acquisition and pretreatment, the feature extraction, classifier design, etc. This paper will use Kinect, a depth of field camera of body sense game machines which is launched by Microsoft and has a very high cost performance, in3D data collection. It partly solved the problem caused by the expensive price and complex operation of the traditional3D data collection equipment. We can also use its perfect function and easy-to-use supporting program interface to pretreatment the3D data such as noise reduction. In the3D face feature extraction, we use kernel principal component analysis (KPCA) method, which projection the3D face data from the high dimension space to low dimensional space keeping the original image in information losses under the principle of less as far as possible, further improve the deficiency of principal component analysis (PCA) in the treatment of the nonlinear data. Based on the strong classification ability and advantage of solving the problem of small sample data, we use the support vector machine (SVM) to classify and recognize the face information data from low dimensional space.In the way of the overall realization of the system, the present study used the depth of field camera Kinect of Microsoft as the hardware device for the collection of face3D data, the Visual Studio2010as the development software tool, simultaneously the MATLAB as the assistant software for simulation experiment. Consequently the study achieved the obtaining, preprocessing, feature extraction and recognition of3D data collection for face.The experiment results show that this method of3D face recognition can Realize3D face data acquisition and processing and assure the good recognition effect, average recognition rate can amount to93.6%. The experimental results show that this research has important significance for the popularize application of3D face recognition technology.