In vivo detection method based on relative attributes

Abstract

The invention relates to an in vivo detection method based on relative attributes. The in vivo detection method comprises the following steps: 1) detecting a human face position on each frame of image in an input video; 2) obtaining human face key points; 3) obtaining an eye or mouth area according to the obtained key points; 4) judging whether the rule of attribute change values of the areas multiple continuous frames of images obtained in the step 3) meets the change rule of a true human face, if so, judging the human face as a true human face, and if not, judging the human face as a false human face. Compared with the prior art, the in vivo detection method provided by invention has the advantages of high detection accuracy, high speed, etc.

Description

technical field [0001] The invention relates to a face detection technology, in particular to a relative attribute-based living body detection method. Background technique [0002] As a kind of identity authentication technology, face recognition has been successful in the fields of public security, attendance and access control. However, traditional face recognition technology does not consider the authenticity of the target face, so it is vulnerable to false face attacks. If the fake face attack is successful, it may cause heavy losses to users, so reliable and efficient face liveness detection technology has become an important part of the face verification system. [0003] Traditional face recognition technology often uses methods such as Fourier spectrum analysis, eye blink detection, and 3D depth estimation to judge living bodies. However, these methods are difficult to meet the requirements of the fields of finance and public security. There are two main factors: 1...

AI Technical Summary

Problems solved by technology

[0004] The performance of traditional eye-closed (or mouth-opened) classifiers is difficult to meet the needs, for example, some people's eyes are small or some people's eyes cannot be completely closed, etc.

Images