Human Face Positive Images Recognition Using Neural Networks Technology

Face recognition (FR) technology is within the research area of pattern recognition, computer vision and image understanding system, face recognition has found a wide range of application in police and security department, such as criminal searching, multimedia monitoring, bank code system, etc. Over the last several years, with the width spread of face recognition technology in economic field, FR arouses unprecedented attentions and become the hotspot in the area of image recognition and understanding.This paper elaborates main theory and technique applied in human face positive images recognition using neural networks, which is composed of three parts: preprocessing, feature extracting, face images classifier. Firstly, in order to eliminate unrelated information in face images, some preprocessing methods are used. When a face image is reading into MATLAB workspace, take some image preprocessing, locate eyes position, extract pure face according to the rule, normalize gray and geometrical dimension, then we can get the normalized pure face images. Secondly, feature extraction employs eigenface method based on K-L transform to extract statistic feature and reduce dimensions. Finally, in the part of face recognition, radial basis function neural networks, as a face classifier, consists of forward part and backward part, adopts an improved learning method which combined linear least square (LLS) and decent gradient method.The algorithms are complimented under the environment of MATLAB 7.0.1 with a face auto recognition system, and train and test the images in the ORL face database. We can conclude that the improved radial basis function neural classifier...