Face Recognition Research And Application In University Enroll System

Face recognition technology as the main content of identity recognition is the hotspot of digital image processing and pattern recognition nowadays.It is used to automatically distinguish identification and solve the problem of ‘who is he’ by the face images collected from computer to extract useful information for recognition.But face recognition acquisition is affected by complex external environment and its own properties such as light,shadow,face-painting,jewelry and headdress.There will be a greater difference in the images obtained from a person,and make it difficult to correctly recognize faces with a machine.Presently,in the field of face recognition,the study keystone is high-precision face image recognition and activity understanding in complex reality environment.High-precision face image recognition researches are focused on image standardization,feature extraction method and face image classification and recognition after face image acquisition.Discrimination decreases sharply when traditional algorithm faced with the complexity and unpredictability of reality environment.In order to overcome these deficiencies,this thesis proposes a adaptive method of the standardization of face image and the feature extraction algorithm of the illumination invariant by the stable characteristic distribution of face image and the prior knowledge of face categorical distribution.Main tasks are as follows:(1)According to Lambertian illumination model,Gamma transform,logarithmic transformation and corresponding histogram equalization.A adaptive model of the standardization of face image is proposed.(2)Based on a deep research into the feature extraction algorithm of the illumination invariant such as LBP,MB-LBP,2D-Gabor and a integration of 2D-Gabor and Uniform LBP,a feature extraction method for reducing characteristic dimension is proposed.(3)We designed a unified standard method for classification.Made classification simulation experiments with the nearest neighbor classification method and SVM and compare the experimental results.(4)Based on Visual Studio 2015 platform,we utilize this algorithm in University Enroll System to recognize the identities of the new students and achieve good results.