Living body identification method and device based on display medium, terminal and readable medium

Abstract

The embodiment of the invention discloses a living body recognition method and device based on a display medium, a terminal and a computer readable medium, and the method comprises the steps: obtaining a to-be-recognized video, carrying out the face recognition, and obtaining a corresponding first face image region; according to a preset external expansion mode, expanding the first face image region outwards in the to-be-identified video, when the edge of the expanded first face image region reaches or exceeds the boundary of the to-be-identified video, taking the boundary of the to-be-identified video image as the boundary of the first face image region, and taking the expanded first face image region as a first target expansion window; obtaining a first feature vector corresponding to the first target expansion window; and inputting the first feature vector into a trained target classifier, and outputting a living body recognition result. By adopting the embodiment of the invention,whether the window containing the human face exists at the boundary corresponding to the display medium or not is detected to judge whether the window is a living body or not, so that the accuracy ofliving body recognition in a human face recognition scene is improved.

Description

technical field [0001] The present invention relates to the technical field of image processing and face recognition, and in particular, to a method, device, terminal and computer-readable medium for living body recognition based on a display medium. Background technique [0002] With the development of computer technology and artificial intelligence technology, face recognition technology has become relatively mature and widely used. For example, in many application scenarios, face recognition is used to identify user identities. Although face recognition technology has high accuracy, there is also a problem, that is, how to detect images that are confirmed as real people, rather than photos, videos or other display media, or deceptive images such as masks and camouflage; How to identify whether the currently detected face is a living body when replacing the actual face with the face in the photo or video. [0003] At present, there are mainly the following methods for liv...

AI Technical Summary

Problems solved by technology

Although the face recognition technology has high accuracy, there is also a problem, that is, how to detect that the image is confirmed as a real person, rather than a photo, video or other display media, or a deceptive image such as a mask; that is, in the user How to identify whether the currently detected face is alive when replacing the actual face with a face in a photo or video

[0003] At present, there are mainly the following methods for live body recognition: the first one, based on actions, requires the user to actively cooperate to make specified actions, such as blinking, shaking the head, etc. This method can prevent photo spoofing, but for video or mask spoofing and other behaviors, this method has less effect; the second, based on the optical flow field, calculates the optical flow field change of the face position through several consecutive frames of images to distinguish between plane and 3D objects, but for deceptive behaviors such as masks, this method The third method is based on color texture, by analyzing the difference between the color texture of the real person and the color texture of photos, videos, masks and other images to confirm whether it is a real person. In theory, this method can identify various forms of deception, however The recognition method based on color texture in the prior art, because it cannot accurately locate the face position, loses color information when extracting LBP (Local Binary Pattern, local binary pattern) features, and does not highlight obvious position features in the calculation, there is also a problem of accurate recognition. low rate problem

[0004] That is to say, in the existing technical solution of living body recognition in the face recognition scene, there is a certain deficiency in the recognition accuracy of living body recognition

Images