Research On Face Recognition Algorithm Based On Gabor And HOG Feature Fusion

Face recognition technology is a technology to extract information from face images and recognize them.Usually,the implementation of face recognition technology needs to go through the process of face feature extraction and feature classification and recognition.How to extract the most effective facial features has always been the focus of academic research.In recent years,more and more feature fusion methods have been proposed.The method based on multi-feature fusion can describe the face image more comprehensively,fully extract the face information in the image,and improve the effect of face recognition.Feature classification is the last stage of face recognition.How to construct a robust and efficient classifier is very important for face recognition task.At present,many classifiers have been proposed.Different feature classification algorithms have different advantages,so we need to choose the appropriate classification algorithm according to the specific application scenarios.Based on this,this paper explores an efficient feature fusion method based on the in-depth study of facial feature extraction algorithm,then makes an in-depth study of KNN(K-Nearest Neighbor)and SVM(Support Vector Machine),which are widely used in the field of face recognition,and explores the optimal classifier through comparative experiments.The main work of this paper is as follows:In order to describe the face image more comprehensively,this paper designs a fusion feature extraction algorithm based on Gabor and HOG(Histogram of Oriented Gradient)feature extraction algorithm.The algorithm first uses multi-scale and multi-directional Gabor filter banks to extract texture features at different levels of the image,filter out irrelevant interference information,and obtain a set of descriptions about the face image,and then use HOG feature extraction algorithm to extract local statistical features for each texture feature image,and obtain the final fusion features.Considering that the dimension of the fusion feature is too large,this paper introduces the Principal Component Analysis(PCA)algorithm to reduce the dimension of the fusion feature.Finally,combines the KNN classifier to verify the effectiveness of the feature fusion algorithm on the ORL and Yale data sets.The optimal parameters of a single feature extraction algorithm are not necessarily suitable for fusion features.In order to explore the expression ability of fusion features,a large number of comparative experiments are designed to explore the optimal parameters,and then the optimal parameters are applied to the fusion feature extraction algorithm.Combined with SVM classifier,a face recognition algorithm is constructed,and the algorithm is tested on a large number of face data sets.