Face recognition method combining fuzzy 2dpca and fuzzy 2dlda

Abstract

The invention discloses a face recognition method combining fuzzy 2DPCA and fuzzy 2DLDA, belonging to the fields of pattern recognition and artificial intelligence. This method first calculates the fuzzy membership value and class center value of the training sample image matrix, and then calculates the fuzzy two-dimensional overall scattering matrix S of the training sample image matrix according to the fuzzy 2DPCA f2DT and S f2DT The eigenvalues ​​and eigenvectors; secondly, according to the fuzzy 2DLDA, calculate the fuzzy two-dimensional inter-class scattering matrix S of the training sample image matrix f2DB , and calculate S f2DT Inverse matrix and S f2DB The eigenvalues ​​and eigenvectors of the product matrix; finally use the eigenvectors of the fuzzy 2DPCA and the eigenvectors of the fuzzy 2DLDA to realize the compression of the face image matrix, and then pull the compressed test sample matrix and training sample matrix into vectors, and The vectors are projected onto the feature transformation matrix, and a nearest neighbor classifier is used to obtain the result. The invention can realize accurate recognition of human face images, and has high recognition rate and high efficiency.

Description

technical field [0001] The invention relates to the fields of pattern recognition and artificial intelligence, in particular to a face recognition method combining fuzzy 2DPCA and fuzzy 2DLDA. Background technique [0002] Face recognition is a hot topic in the field of computer vision and pattern recognition. Two-dimensional principal component analysis (Two-Dimensional Principal Component Analysis, 2DPCA) and two-dimensional linear discriminant analysis (Two-Dimensional Linear Discriminant Analysis, 2DLDA) are commonly used two-dimensional feature extraction methods. The principal components of the face form the eigenface, which can retain the overall spatial information of the sample, but it belongs to an unsupervised learning method, and cannot selectively retain the category information of the sample; 2DLDA uses the image matrix to directly construct the dispersion matrix, and finds the discreteness within the class. The projection matrix with the smallest degree and t...

AI Technical Summary

Problems solved by technology

Two-dimensional principal component analysis (Two-Dimensional Principal Component Analysis, 2DPCA) and two-dimensional linear discriminant analysis (Two-Dimensional Linear Discriminant Analysis, 2DLDA) are commonly used two-dimensional feature extraction methods. The principal components of the face form the eigenface, which can retain the overall spatial information of the sample, but it belongs to an unsupervised learning method, and cannot selectively retain the category information of the sample; 2DLDA uses the image matrix to directly construct the dispersion matrix, and finds the discreteness within the class. The projection matrix with the smallest degree and the largest inter-class dispersion is used to extract features from the data, so it has good category identification ability and is widely used in the field of face recognition.

However, 2DPCA and 2DLDA can only extract features in one direction of the row or column of the image matrix, and cannot extract features in both directions of the row and column at the same time.

In addition, due to the presence of noise information in the process of face image acquisition, 2DPCA and 2DLDA have deficiencies in processing noise information

Images