Study On Face Recognition Algorithm Based On Gray-Value Symmetric Virtual Sample

Face recognition has been concerned by researchers since the middle of the 20th century.The origin of face recognition is early,but it still has the continuing vitality.It is a hot technology that has great appeal to researchers.In recent years,the rise of 5G technology,big data,cloud computing and artificial intelligence has enabled the commercial application of face recognition technology on a large scale.Face recogni-tion technology has a wide range of applications,including video surveillance,suspect tracking,personnel identification,sign-in management,etc.In practical applications,the number of samples training is often limited.One reason is that the storage space of training samples is restricted.Another reason is that it is difficult to obtain a large num-ber of training samples within a short period.At the same time,the face recognition rate with small-scale training samples is relatively low,because of changes in illumination,facial expressions and postures.Firstly,this paper proposes an algorithm using symmetric faces to preprocess the original training samples to generate virtual samples,which makes up for the specif-ic requirements of the variation dictionary for training samples and also enriches the number of training samples and the amount of information on the training samples.The virtual samples are constructed by symmetric faces and then they are substituted into the algorithm framework combining variation dictionary and collaborative representa-tion classification for classification and recognition.The simulation results show that the proposed algorithm still has good classification and recognition performance in the case of small-scale training samples.However,as the symmetric face directly selects the central axis of the image as the location of the symmetrical axis,the algorithm only considers the symmetry of the im-age.Therefore,new virtual sample generation algorithm is proposed to determine the gray symmetry axis which based on the mean of gray values,by analyzing the distribu-tion law of the gray value of the face images.This algorithm can adjust the position of the symmetrical axis in the horizontal direction according to the gray value of the face images.On the basis,virtual samples are generated by the angle variable symmetry axis which is determined by the average gray values of the face images after divided.The proposed algorithm adjusts the inclination angle and direction of the axis of symmetry on the basis of the gray symmetrical axis,which makes the constructed virtual samples have better classification performance.