Implementation Of Face Recognition System Based On Recursive One-class Algorithm

Face is a collection of model contains abundant information, it is one of the main symbol of human mutual dialectical and recognition, it is also one of visual objects of interest in images and video. Face recognition technology combined with digital image processing,computer graphics, pattern recognition, computer vision, artificial neural network, biometric technology and the theory and methods of physiology, psychology, and other disciplines.Facial recognition technology involves image gradually complicated,although face recognition research has had the very big progress, but the technology is still unable to effectively deal with such as faces in complex background and automatic tracking. Therefore,face recognition technology in current and future for a long period of time is still a challenging task.This study aims to discuss the effective method of face detection and recognition.Combining with the characteristics of the current face recognition technology, we put forward an effective method for face recognition. Specifically, this method has four procedures. First,preprocessing collected human face images. Secondly, using morphological operator based face detection algorithm to detect the face region of the pretreatment face images. This algorithm first detects skin color according to the clustering characteristic of the face skin color presented in a certain range. Thirdly, it reuses morphology methods to process the detected face images to obtain face region after the skin color detection. The testing effect of this algorithm is very accurate and the speed of this algorithm is also faster. Second, in the human face feature extraction, the characteristics of human eyes based on Gabor transform is used for facial feature extraction algorithm, this algorithm by writing a Gabor transform kernel function, using Gabor kernel function to realize the Gabor transform of the image, using the expansion method of face image to eliminate some interference region, to get the region of interest and extract the human eye. Fourthly, in the face recognition, this paper adopts a face recognition algorithm based on recursive one-class to solve the dual optimization problems,then gets different characteristic vector, normalize the eigenvector, judge whether the eigenvector meets the termination conditions and take it as the recognition result.On the face recognition method in the MATLAB environment simulation, from the point of the simulation results, face recognition effect is not only effective, but also compared with classical algorithm of face recognition, the face recognition method on recognition rate and recognition time are greatly improved.