Face Recognition Based On Wavelet Transform And HOG Feature

Face recognition is a technology that extracts facial information and uses a classifier to recognize it,which can be used as the unique identification of a person.Face recognition system has the advantages of non-contact,non-encroachment and reliability,so it is widely used in real life,such as high-speed rail inbound,attendance,examinee identification,etc.Face recognition technology is mainly divided into image preprocessing,important feature extraction,feature classification and other links.At present,the algorithms of image preprocessing and feature classification are relatively mature,but there is still room for improvement in the research of feature extraction algorithm.Therefore,this paper studies the face feature extraction algorithm,transplantes the algorithm to the face recognition system,and analyzes its performance.Wavelet transform and HOG are two typical feature extraction methods.Wavelet transform extracts the low-frequency component of the original image with a resolution of one quarter,which filters out high-frequency information such as noise,but contains information similar to the original image.Compared with the original image,the low-frequency component is extracted by wavelet transform contains less interference information.HOG operator can extract image shape information.In order to better describe the shape features of human face,this paper fuses the above two methods and analyzes and verifies the fusion algorithm.The main innovations and specific work contents of this paper are as follows:(1)Explain the wavelet transform and HOG feature principle respectively.Based on the low-frequency feature and HOG feature extracted by wavelet transform,a face recognition algorithm based on wave-hog feature fusion is proposed,and the performance of the classifier is compared.Through the analysis of experimental data,it can be seen that this algorithm has ahigher recognition rate than the existing algorithms.The fusion algorithm proposed in this paper can reach 96.5% and 79.8% respectively under ORL and GT face standard database.(2)Through parameter analysis of the algorithm in this paper,it can be concluded that the optimal parameters of a single algorithm in the database are no longer applicable to the fusion algorithm.According to the experimental data results,the optimal parameters of each algorithm are analyzed and adjusted,and dimensionality reduction is carried out through PCA(principal component analysis algorithm)to test whether the features after dimensionality reduction had shorter recognition time and higher recognition rate.The fusion algorithm proposed in this paper adopts PCA dimensionality reduction in ORL database,and the recognition and classification time is 40 ms,which meets the real-time requirements of the system.(3)In Windows system,open source computer vision library opencv and image user interface toolkit pyqt are used to design the face recognition system.The system includes the functions of image reading and acquisition,image preprocessing,feature extraction algorithm selection and face recognition test.The system is applied to several examples and its performance is tested.The experimental results show that the system has good face recognition ability.