The Research On Face Recognition Approaches Of Infrared Imagery

Facial representation and recognition is one of the key issues in computer vision , pattern recognition and biometrics fields.lt is widely applied to high safety section's surveillance,entrance controls and computer security system and so on.Recent researches are focused on visible image face recognition with frontal or frontal plus profile in a simple background,but infrared image have a lot of advantages such as strong anti- interference , independence to visible light source, defendence from camouflage and defendence against cheat etc.,these advantages making infrared image face recognition can make up to a large extent visible image face recognition's shortage.Now infrared image face recognition is an important direction of face recognition.According to the actual application and based on the analysis of characteristics of infrared face image,the technologies and the methods of infrared image face recognition are stutied systematically in this paper.According to the mechanism of infrared imagery,the properties of infrared face image,the main features of infrared image face recognition,factors that influence infrared image face recognition and the characteristics of infrared image face recognition are analyzed which provides the foundation for the study on infrared image face recognition.The image preprocessing method of infrared image face recognition is studied based on the analysis of properties of infrared face image.The principles of the Gauss differential calculus and the symmetry transformation are analyzed;The technology of the face location using Principal Component Analysis( PCA) template match method is studied and a new method for infrared image face location is developed,which is according to symmetry and Eigen-face template match.The new method is stability and has stronger practical value.Based on the analysis of statistical face recognition method and factors that influence infrared image face recognition,we propose an excellent infrared image statistical face recognition method; According to the shortages of statistical face recognition method in infrared image face recognition field,a new Linear Discriminant Analysis(LDA) infrared image face recognition method and an improved Independent Component Analysis(ICA) method are proposed in this paper.These new methods are adaptive to infrared image face recognition and their performance are good and have the certain theoretic significance with the actual application value.According to subspace pattern recognition and the non-linear face image analysis method,a novel method called non-linear feature subspace for infrared image facerecognition is proposed in this paper.The ill-conditioned problem caused by small sample size and often encountered in other statistical face recognition methods is avoided in this method. This new method is effective in the case of small sample size problem in infrared image face recognition tasks.According to the properties of the features which obtained by some statistical face recognition methods such as PCA,LDA and ICA etc. are different,a new decision fusion infrared image statistical face recognition method which combine the results of the statistical face recognition methods is proposed;Based on the analysis of the respective properties of infrared and visible image face recognition and the studies of image fusion technique,we propose a new image fusion face recognition approach that combining infrared and visible image face recognition,which is viable theoretically and can solve problems in actual application such as variational illumination,face pose changes, face decorates, facial expression changes , the quality of infrared face images is not good ,wearing glasses and between the gallery and probe image acquisitions having time lapse etc..