Research On Feature Extraction And Matching Algorithm Of Face Recognition

Face recognition technology is a set of cross subjects in pattern recognition,image processing,machine vision,human psychology and artificial intelligence technology,and according the exploring of researchers,it has gained a lot of impressive results.Compared with other biometric identification technologies,this technology has many advantages,such as non-contact operation,low cost,good scalability and so on.The general face recognition system can be divided into four parts: image preprocessing,face detection,feature extraction and classifier design.Among them,feature extraction is the most important step in face recognition,and its quality directly affects the recognition rate of face recognition.In this paper,the feature extraction algorithm is the main research item,whose background is face recognition and image matching.And this paper proposes improved feature extraction algorithm for different situation and classifiers,and verifies the feasibility and efficiency of the algorithm.The main contents are given as follows:Firstly,because of the confusion of color information and the low value of the edge response in the collected photos,we propose to do the grayscale processing and edge detection of the photos.Gaussian filter is chosen as the smoothing of picture pretreatment to to eliminate the image noise spot.Aiming at the limitations of SIFT algorithm and SURF algorithm,which are the main feature extraction methods in face recognition,a feature extraction method based on the combination of SURF operator and spatial Pyramid expression is proposed,which not only guarantees the extraction of the local features of the face,but also shows the global characterization of the whole image,and achieves the local global feature extraction of face recognition.Then,the SVM and Adaboost classifiers are selected to classify the extracted facial features,and the performance advantages of the algorithm are verified.According to the problem of high degree of exposure,fuzzy picture,angle torsion and noise problem of image,the image matching experiment is designed.The FLANN method is used as the matching strategy,and the matching performance is verified by the proposed method based on SURF operator and space Pyramid.