| Video based face recognition is one of the important issues in computer vision and pattern recognition, as it has great research value and potential market applications, in recent years, it attracts many scholars and research institutions to engage in the research field. At present, the recognition technology has achieved a certain development, but because of the complexity of human face patterns, face recognition is still confronted with many difficulties and challenges, to achieve a relatively mature technology and widely applied to practice, there are a lot of work to be done.In this paper, face recognition based on video surveillance is the main line, following the main line, face detection, preprocessing, feature extraction and classification are studied, a face recognition system based on video surveillance is designed and implemented. In image acquisition, a video capture frame with Microsoft's digital video software development kit VFW is bulit, through the camera video images are captured. In face detection, the commonly used methods of face detection are studied, from the consideration of the detection rate and real-time, AdaBoost algorithm based on cascade classifier is trained and used to detect faces from the video. Meanwhile, in order to improve the image quality, the pretreatment such as the noise remove, geometric normalization and intensity normalization are used to the detected human face. In face recognition, the principal component analysis is in-depth studied, and is applied to face recognition, it is used to extract facial feature, and reduce the dimension of facial features, construct subspace, then the face from the video is recognized by the nearest neighbor classifier with traind templates using a pre-determined strategy. At last, relevant data are collected to test the system, experimental results show that the system in terms of face detection and recognition has a better recognition rate, and can meet the requirements of real-time system. |
PDF Full Text Request