Face Recognition Method Based On MLBP-HOG Combined With Gray Level Co-occurrence Matrix Features

Face recognition is a biometric technology based on human facial feature information for identity recognition.It is reliable,non-contact,and fast.It is used in various fields of life.Face recognition technology has developed very mature,but affected by factors such as illumination,posture,and facial expression,face recognition technology still has many difficulties that need to be researched and solved.This article focuses on the feature extraction part of the face recognition process.In order to better describe the image features,this article describes the local details of the image,the LBP algorithm can better describe the local texture features of the image,and has good robustness to different poses and lighting;The HOG feature can describe the edge gradient information of the image,and has a certain degree of invariance to geometric and optical changes;The characteristics of the gray level co-occurrence matrix describe the spatial correlation characteristics of pixel values.This paper integrates and improves the three characteristics.The main work and innovations are as follows:(1)In order to solve the problem that LBP is easy to lose some details when extracting facial features,an improved LBP algorithm based on the most value averaging is proposed(MLBP).This method calculates the nine pixel variance for a 3×3 template,if the variance is within the limited range,the average of the maximum and minimum values of the eight pixels around the center pixel is taken as a threshold value,avoid the fact that the central pixel value is too large or too small to ruin the details,thus retaining more local details,otherwise,the median of nine pixels is used as a threshold for comparison to reduce noise.(2)In order to preserve the edge gradient characteristics of the image and the spatial correlation characteristics of pixel values,this paper proposes the MLBP-HOG-G algorithm.Use the MLBP algorithm to extract binary texture images,and then use the HOG algorithm to perform secondary feature extraction on the binary texture images to obtain MLBP-HOG features,the PCA is used to reduce the dimension of the MLBP-HOG feature vector,and then the dimensionally reduced MLBP-HOG feature and the gray level co-occurrence matrix feature weight are serially fused to obtain the fusion feature MLBPHOG-G.(3)In order to construct a face recognition system,this paper cascades the classifiers SVM and KNN,selects SVM linear separability and KNN classification,constructs a cascade classifier,and uses this to combine feature extraction to construct a face recognition system.Using MATLAB to experimentally verify the improved algorithm designed in this paper and the face recognition system algorithm,through the comparison of experimental results,we can conclude that in the experimental sample space of this paper,the improved algorithm proposed in this paper has a certain improvement in recognition rate,therefore,the experimental results in this paper verify the effectiveness of the improved algorithm in this paper.In addition,a face recognition system based on the GUI tool in MATLAB was produced.This system can complete simple face identity information verification.