Near-infrared Face Recognition Algorithm

With the development of information society, more and more occasions need rapid and effective identity authentication, such as bank systems,customs,network account login,access control etc..Compared to the password, door locks that are traditional identity authentication methods and other biometric identification technology, face recognition as one of biometric identification methods has the following advantages:①more security,non-invasiveness,prevention of amnesia and infringement;②substantivity, friendship, convenience, acceptably for users;③friendly operation, good concealment, intuitive, simplicity of acquisition equipment.So face recognition technology has been widely studied and applied in many areas, and has become a research hotspot in the field of biological recognition.After several decades of development, face recognition technology has made great progress. So far, face recognition system based on visible light get the greatest use, this kind of systems are easily affected by a series of external factors, such as illumination, pose, which limit its development in the field of face recognition.In recent years, In the light of existing faults of the visible face recognition, some scholars put forward a kind of new face recognition technology in which active near infrared is used.This method reduces effects of light which is unfavorable factor, and also improves adaptability of face recognition algorithm when Illumination changes.This face recognition system can works even at night or in dark, thus this technology has become a hot topic in face recognition field in recent years. The main research contents of this paper are as follows:Firstly, this paper analyzes the research background and current situation of the development of face recognition technology at home and abroad; and then makes a deep analysis of the existing face recognition techniques (including visible face recognition, thermal infrared face recognition and near infrared face recognition which the paper mainly focuses on).The vast majority of face recognition systems generally use the visible face images to recognize, such systems are vulnerable to the impact of environmental light. In general, before distinguishing we often make some preprocessings of light interference which although can eliminate the influence of light in a certain extent, but also lose some useful information.In order to solve the impact of illumination, pose, expression and other factors on the performance of face recognition, this paper proposes a near infrared face recognition scheme.Secondly, this paper makes classification and summary of the existing face recognition methods, and introduces the processes of face recognition systems (including two aspects of the general processes and specific process), and makes a detailed introduction of each step of the processes,later makes a brief introduction of the products of classic and practical face recognition systems at home and abroad, which further proves the face recognition technology has penetrated into various areas of life and produce.finally the article briefly introduces the indexes of evaluating face recognition system performance, and explains the meaning of each index.Finally, this paper analyzes the difficulties of face recognition method,and make further improvement of the existing face recognition methods.In response to the problem that the existing near infrared(NIR) face recognition methods yield insufficient robustness to the varieties of facial gesture and expression, this paper describes a new algorithm for near infrared(NIR) face recognition based on Contourlet transform,Non-negative Matrix Factorization (NMF) and support vector machine (SVM).The algorithm divides the whole recognition process into two steps of feature extraction and matching.In the step of feature extraction, a NIR face image is decomposed into several sub-images via Contourlet transform, and then NMF is utilized to decompose each sub-image into a U matrix and a V matrix. The first-order statistics of the V matrixes are determined and regarded as the features.Then, in the step of matching, on the basis of selecting a particular database a support vector machine-based classifier is utilized to evaluate the algorithm of recognition performance.The algorithm effectively combines the feature extraction of Contourlet transform,the dimension reduction of NMF and the classification ability of SVM.Experimental results show that the proposed approach yields a better performance in term of the correct classification percentages compared with the recent NIR face recognitions.It is also shown that the proposed approach yields observably high robustness to the varieties of facial expression and gesture.