A face recognition method and system

Abstract

The invention discloses a face recognition method and system, which includes extracting feature vectors of labeled and unlabeled human face images to obtain a first feature vector set and a second feature vector set; calculating soft labels of unlabeled human face images Information; calculate the inter-class scatter and intra-class scatter of the face image training set and execute the linear discriminant analysis algorithm to obtain the discriminant projection matrix; use the discriminant projection matrix to test the eigenvectors of the face image, the first eigenvector set and the second Two feature vector sets are subjected to dimensionality reduction to obtain respective low-dimensional feature vectors, and the low-dimensional feature vectors are input into a collaborative representation classifier to obtain respective collaborative representation codes, and the reconstruction residuals are calculated using the collaborative representation codes corresponding to each class, The label with the smallest reconstruction residual is the sample label to be tested. The present invention transfers label information of labeled data to unlabeled data, expands the number of labeled training samples, and conducts identification analysis on all samples, thereby improving the accuracy and discrimination of CRC.

Description

technical field [0001] The invention belongs to the field of face recognition, and in particular relates to a face recognition method and system. Background technique [0002] Face recognition is one of the most important applications in the field of pattern recognition technology. The so-called face recognition is to use a computer to analyze the face video or image, and extract effective personal identification information from it, and finally determine the identity of the face object. [0003] Collaborative Representation Classification (CRC) is an efficient and high-speed classifier, which has been widely used in the field of image recognition. The basic idea of ​​CRC is to use all the training samples to jointly represent the test sample, and then obtain the collaborative representation code. Then, a class code selector is used to select the class code corresponding to a certain class of training samples, and finally, the class code is used to calculate the reconstruc...

AI Technical Summary

Problems solved by technology

[0004] CRC is essentially a supervised learning method, and its performance depends heavily on the number of labeled training samples of each type. When the number of labeled training samples of each type is not sufficient, the recognition performance of CRC will be significantly reduced.

[0005] Secondly, although CRC has a good representation ability, its ability to distinguish different types of samples is weak. In order to increase the robustness of the model and make the face images of different people distinguish as much as possible, it is necessary to increase the discrimination ability of the CRC model itself.

Images