Research And Implementation Of The Authentication System Based On Face Recognition

Face Recognition technology is an important research area of computer vision and artificial intelligence. In the last several decades, artificial intelligence research is developing rapidly and achieves important results. However, in actual video face recognition, existing techniques are far from satisfaction because the recognition accuracy is still interfered by background, light, expression, shade, “fake face”, etc.To address these problems in actual video detection environments, this thesis improves existing approaches and proposes algorithm to achieve face detection and recognition system as follows:(1) To eliminate interference of the background image for face detection in video, this thesis proposes a method to use a custom background model for video image background segmentation, so as to narrow the scope of the face detection in video images.(2) In order to reduce the influence of light, backlighting and other image noise on facial feature extraction, this thesis proposes histogram equalization method to pre-process all pending images. According to the new approach, the recognition degree of human facial detail features can be significantly improved.(3) To reduce the error rate of “fake face” detection, this thesis proposes a skin color-based model in RGB color space to further analyze the correctness of face detection, so that the suspected face can be determined.(4) In the facial feature recognition, this thesis uses iteration penalty strategy to calculate the face LBPH feature trainer recognition rate and error rate for crossing training set, and gets an adaptive confidence threshold for facial feature matching, recognition, and improving the system's robustness.This thesis has implemented the authentication system based on face recognition, and has carried out in depth research on Haar facial feature extraction, AdaBoost Cascade training, and LBPH facial feature. Based on the prototype, we conduct extensive experiments to demonstrate the feasibility of the approaches and improvements presented above. The study in my thesis has a positive meaning on face recognition technology research.