A two-stage face image classification method and system under small sample conditions

Abstract

The invention discloses a two-stage face image classification method and system under the condition of a small sample size, wherein the method includes: a sample expansion stage, which uses a semi-supervised method to make the unmarked samples of the face image cooperatively represent the marked samples to obtain a collaborative Indicate the coefficient; obtain the unlabeled sample corresponding to the maximum synergy representation coefficient; add the unlabeled sample corresponding to the maximum synergy representation coefficient to the marked subset, expand the marked sample, and use the expanded marked subset as the training sample; at the same time Use the remaining unlabeled samples as new unlabeled samples; in the face image classification stage, based on the collaborative representation classifier, use the expanded label subset to classify the new unlabeled samples obtained in the sample expansion stage, and obtain the final classification results. The invention improves the accuracy rate of the supervised classification method, and at the same time makes full use of the judgment information of unmarked samples, and converts the semi-supervised learning problem into a supervised learning problem by using a sample expansion method.

Description

technical field [0001] The invention relates to the field of pattern recognition, and more specifically, relates to a two-stage face image classification method and system under the condition of small samples. Background technique [0002] With the rapid development of pattern recognition and computer vision technology, face recognition has attracted the attention of many researchers from various fields due to its wide application, and has become an important aspect of modern pattern recognition technology research. [0003] However, face recognition is a small sample problem in practical applications. Many traditional face recognition methods are based on a large number of training samples. Therefore, in the case of an extreme lack of labeled samples, a large number of supervised recognition methods are limited. Recognition ability will be weakened. At present, the main face recognition methods used include KNN (k nearest neighbor), LDA (linear discriminant analysis), SRC ...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to solve the classification problem of a large number of unlabeled samples under the condition of small samples, and to provide a two-stage face image classification method and system under the condition of small samples. The advantages of achieving efficient face image classification under the condition of samples

Images