Face recognition method and device

Abstract

The invention relates to a face recognition method and device. The method comprises the following steps: acquiring a first image and a second image which comprise the face of an object to be tested, wherein imaging methods of the first image and the second image are different; extracting first feature information from the first image; extracting second feature information from the second image; carrying out data fusion on the first feature information and the second feature information; carrying out optical flow processing on the feature information after data fusion; and judging whether the object to be tested is a living body or not according to the feature information processed by optical flow. Features of images acquired by the different imaging methods are subjected to data fusion, and then are processed by the optical flow, whether the object to be tested is a living body or not can be determined, therefore, replaying cheating caused by using photos or videos is prevented, and the invasion resistance of face recognition is improved. In addition, features are weighted by various feature parts of the face, and the recognition speed, invasion resistance and light adaptation of infrared human face recognition can be improved further.

Description

technical field [0001] The present disclosure relates to the technical field of image processing, and in particular, to a face recognition method and device. Background technique [0002] Face recognition is a very hot research area in biometric authentication. Face recognition is favored by people due to its natural interactive interface, cheap acquisition equipment, non-contact acquisition, and ease of use. It can be used in access control and attendance, system login, immigration management, and suspect photo comparison. However, there are not many researches on the detection of face counterfeiting attacks, which restricts the application of face authentication systems. At the same time, the diversification of face impersonation methods also brings great challenges to the research of face impersonation attack detection technology. Therefore, the research of face impersonation attack detection technology has increasingly become a hot spot in the field of biometric authent...

AI Technical Summary

Problems solved by technology

[0007] 2. Poor security

At present, the infrared face recognition system has certain advantages for static images, but it is vulnerable to external video or image replay, 3D model attacks

For example, experiments by Kollreider et al. showed that the optical flow method can effectively detect photo spoofing, but cannot detect video spoofing.

This method also requires the user's head to make a little movement, and because the method estimates the movement of different positions of the head through optical flow, the change of light will seriously interfere with the method's estimation of the movement of different parts

[0008] 3. Weak adaptability

Part of the face recognition method is based on image quality and reflectivity, so the image has high requirements on the lighting of the application scene, which limits the application of face recognition

For example, the high-frequency descriptor proposed by J. Li et al. is mainly based on the fact that the clarity of the face will generally decrease after the photo is imaged. Therefore, if a high-resolution photo of a large face is used to deceive, this method cannot be detected.

Images