Based On The Research Of Aam Facial Modeling And Matching Technology

In the information technology era, people have no chance but no confront the explosion of knowledge and the security of data. Then, biometric feature identification technology has been more and more attentioned, and face recognition becomes one of the important research field of biometrics identification technology with its many advantage. The complete process of face recognition mainly includes four aspects that is the face images capture, face detection, facial feature extraction and location and facical feature matching. In the process, facial feature extraction and feature matching play a so decisive role that direct impact on the accuracy and efficiency of face recognition. In this paper, we have researched the classcal Active Appearance Model and the Inverse Composition image ilignment Thoroughly, and to solve the problem of illumination, angle, expression and computing time complexity, we propose the Semi-Active Appearance Model that reduce the time complexity and space complexity of data computing and improve the accuracy of matching and recognition in the process of modeling and matching.The main research and achievements of this paper are as following:(1) Studied the classcal Active Appearance Model and its related matching calculation. The Active Appearance Model consists of the Shape Model, Texture Model and the Combination Model. This model has the advantages of the accurated face feature localization and the good representation of the appearance of face images, and the disadvantage of the complex calculation and the low computational efficiency because of using the method of traditional gradient descent method.(2) Studied the Inverse Compositional Image Alignment, owing to the lower efficieny of classical gradient descent method in the matching calculation. The method was proved by the Lucas-kanade algorithm and the forward Compositional Image Alignment. We improved the computing efficiency in the matching calculation by the Iverse Compositional Image Alignment.(3) Proposed the Semi-Active Appearance Model to solve the problem of illumination, angle, expression and computing time complexity. The model got from the combination of Active Appearance Model and Grey Level Co-occurrence Matrix. We eliminated the calculation of engenvalues and eigenvectors in the process of establishing the model, and used the Inverse Compositional Image Alignment to the process of matching calculation in order to improve computational efficiency and recognition accuracy. We proved that the method had a simple calculation and higher accuracy of identification by some experiment, and so that solved the problem of Active Appearance Model.(4) Achieved the complete face recognition software based OpenCV and MFC. The complete process consists of four parts that is image capture and face detection, feature extraction and location, feature matching and recognition and data saving.