The Research On Face Detection And Recognition Based On Real-time Video

With the developments of machine vision and human-machine interaction, biometric recognition has become a popular research area in the pattern recognition and image analysis societies. Since human face detection and recognition system could gain target indirectly and secretly, it has attached more and more importance. Besides, human face information could be acquired easily and hidden hardly, it is convenient for surveillance in public; at last, face detection and recognition system could be interacted with each other momentarily and helpful in investigation and tracing.Just as its name implies, human face detection is the process to judge whether the human face exists in a certain condition and to make sure the number, location or any other available parameters. Face recognition is the process to distinguish the identity of the certain person by comparing the detected face and the gathered databases. The research project of this paper aims at the real-time application of face detection and recognition system in small-scale situations, the main task is as followed:1. A multi-information inosculation method including motion, color and transcendent information for face detection and confirm is proposed. The interested targets would be detected first by motion information; and then the improved skin color model in YCbCr space will be employed to select the parts that may be skin such as face, neck, hands and so on; the transcendent information will be utilized by support vector machine classifier to ensure the face parts.2. A novel face recognition method based on principal component analysis and kernel partial least squares for feature extraction and support vector machine for classification is presented after the analysis of some feature extraction methods, and compared with principal component analysis, singular value decomposition, linear discriminant analysis and partial least squares by using k-nearest neighbor and support vector machine classifier. Experiments on public database have proved the effectiveness of the presented method.3. The DirectShow was introduced for USB camera drive. It is part of Microsoft open source development kit. Face detection and recognition software could catch and replay multimedia streams conveniently, and be hardware irrespective and accessing real-time by using DirectShow.