Face recognition method based on channel feature fusion sparse representation

Abstract

The invention discloses a face recognition method based on channel feature fusion sparse representation. The face recognition method is low in calculation complexity, is high in operation speed, can conveniently solve the problem that categories are increased or decreased in practical application of face recognition, and has relatively high robustness for various changes of a human face in a realliving environment, such as the shooting angle, changes of expressions and the like. The face recognition method comprises the following steps: removing a full connection layer of a pre-trained ResNetnetwork, wherein the rest part is used as a feature extraction network; inputting a to-be-detected picture and a picture in the sample library into the feature extraction network, to obtain feature maps thereof; adding the feature maps of every 64 channels to obtain fusion feature maps of eight channels; performing sparse representation classification on the eight fusion feature maps of the to-be-tested image and the picture in the sample library, to solve sparse representation coefficients of the eight fusion feature maps; and finally adding category difference values solved by the eight channels, wherein the category with the minimum difference value is the category of the to-be-tested face image.

Description

technical field [0001] The invention relates to the technical field of face recognition, in particular to a face recognition method based on channel feature fusion and sparse representation. Background technique [0002] Face recognition technology has developed rapidly in recent years and has been widely used in many practical scenarios, such as mobile phone unlocking, immigration management, criminal investigation, etc. The current mainstream face recognition technology is mainly divided into two categories, including traditional methods based on statistical learning, such as PCA, LDA, sparse representation, etc., and methods based on deep convolutional neural networks. The deep convolutional neural network has a strong nonlinear expressive ability, so it works well when applied to deep learning. However, most current face recognition methods based on deep learning have a problem, that is, the classification categories are fixed. If you want to increase or decrease a clas...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a face recognition method based on channel feature fusion sparse representation, to solve the problems of complex calculation and slow operation speed in most of the face recognition algorithms with high recognition rate proposed in the background technology

Images