Image recognition method and device based on sparse marginal Fisher algorithm

Abstract

The invention discloses an image recognition method and device based on a sparse marginal Fisher algorithm. The method comprises the following steps: obtaining a first training data set and obtaining a projection matrix according to the first training data set; sparsifying the projection matrix to obtain a sparse projection matrix; projecting the first training data set through the sparse projection matrix to obtain a second training data set; receiving first test data of a to-be-recognized image, and projecting the first test data through the sparse projection matrix to obtain second test data; recognizing the second test data on the second training data set with a classification algorithm. Therefore, the projection matrix is obtained by projecting the training data set, geometric structure of data can be kept to the greatest extent, then the projection matrix is sparsified, better generalization ability and discrimination ability can be obtained, the problem of balance loss of discriminative learning of all data is solved, and the recognition rate is increased.

Description

Technical field [0001] The invention relates to the field of image recognition, and more specifically, to an image recognition method and device based on a sparse boundary Fisher algorithm. Background technique [0002] With the development of artificial intelligence technology, the forms of human-computer interaction continue to expand and enrich, and the construction of effective human-computer interaction has become a development trend. Face is the most important and direct way of human emotion expression and communication. Therefore, face recognition technology has gradually become a hot spot in the field of pattern recognition in human-computer interaction. Recognition methods based on facial features have gradually developed, but the methods used to extract facial features are not yet mature and reliable. [0003] The Boundary Fisher Analysis (MFA) algorithm is a feature extraction method. It is a graph embedding algorithm that combines class boundary information. The basic ...

AI Technical Summary

Problems solved by technology

The recognition method based on facial features has gradually developed, but the extraction method of facial features is not mature enough.

[0003] Boundary Fisher Analysis (MFA) algorithm is a feature extraction method. It is a graph embedding algorithm that combines class boundary information. It makes the boundary sample points between different classes more separated, makes full use of the inter-class boundary discrimination information and the local similarity information within the class, and preserves the geometric structure of the data well, but due to the local discrimination structure, the global discrimination information of the data set is ignored , which may lead to unbalanced discriminative learning for all data

Images