Improved PCA face recognition algorithm capable of resisting illumination influence

Abstract

The invention belongs to the field of image processing, and discloses an improved PCA (Principal Component Analysis) face recognition algorithm capable of resisting illumination influence, which comprises the following steps of: (1) selecting a training sample set; (2) classifying according to the angle and illumination of the image, and dividing the image data into a plurality of sub-data sets; (3) calculating a mean vector, a centralized data matrix and a covariance matrix of all the sub-data sets; (4) calculating eigenvalues of the covariance matrix, selecting k maximum eigenvalues from theeigenvalues, and solving eigenvectors corresponding to the k maximum eigenvalues; adding a weight coefficient to each feature vector, and arranging the feature vectors into a transformation matrix Waccording to columns; (5) calculating projection matrixes of all images in the training sample set and storing the projection matrixes; and (6) calculating a projection matrix of a to-be-recognized face, and traversing projection matrixes of all head portraits in the search sample training set for matching calculation to obtain a matching result. The method can reduce the illumination influence, and improves the recognition rate.

Description

Technical field [0001] The invention belongs to the field of image processing, and in particular relates to an improved PCA face recognition algorithm resistant to the influence of illumination. Background technique [0002] PCA (Principal Component Analysis), also known as principal component analysis, is a commonly used effective method for processing, compressing and extracting information based on the variable covariance matrix. Its basic principle is to use K-L transformation to extract the main components of the face to form a characteristic face space. During recognition, the test image is projected into this space to obtain a set of projection coefficients, which are then compared with each face image for recognition. [0003] The K-L transformation uses an orthogonal matrix composed of the normalized orthogonal eigenvectors of the covariance matrix of the original data as the transformation matrix to achieve data compression in the transformation domain. It has the charac...

AI Technical Summary

Problems solved by technology

[0005] The realization of traditional PCA requires a lot of assumptions in theory, which makes it impossible to effectively identify theoretically in many cases.

First of all, the traditional PCA algorithm requires the standard training matrix to conform to the Gaussian distribution. When the probability distribution of the investigated data does not satisfy the Gaussian distribution, the variance and covariance cannot be used to properly describe the noise and redundancy, and the training space cannot be well reflected. The feature subspace of , which will inevitably make the recognition rate of PCA relatively low

[0006] 1. The probability distribution of the investigated data does not satisfy the Gaussian distribution and cannot resist the influence of noise;

[0007] 2. The image is greatly affected by illumination changes. Traditional PCA does not consider the influence of illumination, and the weight of the feature vector is the same

Images