A recognition method, recognition device and electronic equipment for face authenticity

Abstract

The application provides a recognition method, recognition device and electronic equipment for the authenticity of a human face. Pixel regions with different sizes and types are determined on the face image to be recognized; The number of pixel areas under the size is spliced ​​to determine the spatial feature map; for the spatial feature map corresponding to the pixel area of ​​​​each size, each spatial feature map is input to the trained multi-convolution cascade, and the output of the size is The spatial feature vector corresponding to the pixel area of ​​, wherein the multi-convolution cascade device is composed of multiple convolutional layers and multiple skew convolution layers alternately connected; according to the spatial feature vector corresponding to the pixel area of ​​​​multiple sizes and types, determine Face authenticity recognition results. This application processes pixel areas of different positions and sizes sequentially through multi-level and multi-level traditional convolutional layers and skew convolutional layers to automatically obtain the recognition results of the authenticity of the face, which has a high recognition accuracy. .

Description

technical field [0001] The present application relates to the technical field of identity verification, and in particular to a recognition method, recognition device and electronic equipment for the authenticity of a human face. Background technique [0002] At present, identity verification methods using face recognition can be seen everywhere, but the security problems that come with it are also emerging in an endless stream. People have gradually realized the defects of face recognition technology. For example, the attacks on face recognition system can be forged by using printed paper, mobile phone screen photos or masks. In order to prevent the above attacks, reduce the number of face recognition systems If the recognition error rate is high, it is necessary to use face recognition technology to identify various attack behaviors. [0003] Existing face detection technology relies on artificially set image features with forgery characteristics: such as recognition of th...

AI Technical Summary

Problems solved by technology

[0003] Existing face detection technology relies on artificially set image features with forgery characteristics: such as recognition of the border and frame of printing paper, reflection of mobile phone screen, Fourier spectrum, recognition of black border of mobile phone and color of mask As well as reflected light recognition, a neural network is constructed to identify and process these features through preset features, but the recognition accuracy of these methods is low, and there are still potential safety hazards

Images