Research On Key Algorithms Of Face Recognition

At the moment,face recognition has been one of hottest AI technology,which is playing an increasingly important role in identity authentication.In this paper,some important algorithms related to face recognition are studied to meet requirements of intelligent access control.This paper introduces two algorithms of face detection,which are based on Adaboost algorithm and HOG-SVM algorithm.By comparing the two algorithms,it is found that Adaboost algorithm detects faces faster,while HOG-SVM algorithm detects faces more precisely with longer time.Some image preprocessing methods on illumination can enhance details of face images before recognition.This paper mainly focuses on Eigenface,Fisherface and local binary pattern histograms(LBPH)which are based on features of images,and takes recognition rate and time as the evaluation criteria to evaluate the three algorithms.It is found that LBPH is robust to illumination but recognizes faces slowly.Eigenface is sensitive to illumination.The recognition rate of Fisherface can be largely improved if images are preprocessed.Finally,it is found that only when there are few face-shielding objects and the face posture is relatively correct,the requirements of low false acceptance rate and low false rejection rate can be met.On this basis,this paper further reduces the face area of images,and proposes a recognition algorithm based on LBP operator combined with Eigenface and Fisherface,which keeps low false acceptance rate and false rejection rate of the intelligent access control system at the same time.If appropriate LBP operation is selected,it can be ensured that both rates can reach 0,which fully meets the security requirements of intelligent access control.