Over-age face verification method based on deep learning and dictionary representation

Abstract

The invention discloses an over-age face verification method based on deep learning and dictionary representation. The over-age face verification method comprises the steps of performing key point calibration on a to-be-verified face image for obtaining key points of the face; extracting a local area face block which corresponds with each key point, and obtaining a local face block which corresponds with each key point; inputting the local face block into a trained deep convolutional neural network, extracting high-layer characteristics of the local face blocks, wherein one multidirectional vector which represents the high-layer characteristic of the face block can be obtained from each face block; acquiring a plurality of images, performing operations on outer data, extracting the characteristic of each age of each area of each kind, thereby forming an outer data reference set; calculating coding vectors of a training image and a testing image in the outer data reference set; and according to the coding vectors of each sub-block in the outer data reference set, obtaining a correct identification result by means of minimal sum of cosine similarity.

Description

Technical field [0001] The present invention relates to the field of computer vision, and more specifically, to a cross-age face verification method based on deep learning and dictionary representation. Background technique [0002] The development of science and technology has made camera equipment popularized, and a huge amount of face image data has also been produced. At the same time, face verification technology needs to be applied in many fields, such as entrances to various venues, customs transit passages, etc. In these applications, only two cross-age face images may be obtained, which creates the problem of cross-age face verification. Almost all the applications of face verification technologies are based on the same age group, but once it is necessary to verify two cross-age face images, these face recognition technologies and systems are at a loss and cannot be applied. Therefore, by solving the problem of cross-age face verification, the application range of face...

AI Technical Summary

Problems solved by technology

[0005] Most of the above methods cannot achieve good results in cross-age face recognition, and these methods cannot be applied to cross-age face verification occasions

Images