Research On Face Recognition Method Combining Global And Local Features

In the past 30 years of scientific and technological development,the research of face recognition has been favored by a large number of researchers.The key to face recognition research is the effective extraction of facial features.The facial features can be divided into local features and global features.The algorithm based on face local feature extraction has local binary pattern(LBP)and scale invariant feature transform(SIFT).and many more.The face recognition method based on local binary mode is robust to changes such as illumination and rotation,mainly to extract texture features of images.Scale-invariant feature transform(SIFT)features are mainly based on some scales and scales on images.A stable point that is independent of illumination,illumination,and rotation,which can robustly identify objects between clutter and partial occlusion.Based on the principal component analysis(PCA)and two-dimensional principal component analysis(2DPCA),the principal component analysis(PCA)and two-dimensional principal component analysis(2DPCA)are used to find the face expression of the face image in its feature space.Based on the above algorithm,this paper proposes a face local feature extraction method based on the combination of SIFT and MTLBP(Mul-threshold LBP)algorithm for the singularity of traditional algorithm in extracting facial features and LBP algorithm for extracting facial texture features.And combined with the 2DPCA algorithm to extract the global features of the face to complete the expression of the entire face feature information.The main research contents are as follows:(1)The Adaboost algorithm for face detection is studied,and the facial features of the face are extracted by the Adaboost algorithm.In terms of image preprocessing,grayscale transformation,histogram equalization and image filtering are performed on each face image sample.(2)For the extraction of local features of face images,there is no specific problem for LBP features in face texture feature extraction.Firstly,the face is sent to the five senses by Adaboost algorithm,and the facial features are rotated by SIFT algorithm.The invariable key points are extracted,and finally the SIFT algorithm and the MTLBP algorithm are combined to extract the local features of the face.(3)Extracting the global features of the face,using PCA and 2DPCA algorithm to extract the global features of the face,using the more efficient 2DPCA algorithm to extract the global features of the face and merge with the local features to form a complete person.Face information feature expression.(4)In the experimental part,the fusion algorithm of local features and the fusion algorithm of local features and global features are discussed experimentally,and the nearest neighbor classifier is used for classification.Finally,the experimental data is compared with the face recognition methods of other literatures.