Face representation attack detection method and system based on fusion features and dictionary learning

Abstract

The invention discloses a face representation attack detection method and system based on fusion features and dictionary learning. The method comprises the steps of extracting image quality features of a complete face image according to a distortion source of secondary imaging of the face image; constructing a deep convolutional network model, and extracting deep network features of the face imageblocks through a deep convolutional network; cascading the two features to generate a final fusion feature through PCA; initializing dictionary atoms by using the fusion features, and training a dictionary learning classifier based on a low-rank shared dictionary; and judging the category of the test sample based on the size of the fusion feature reconstruction residual. According to the method,face representation attack detection is carried out by combining image quality features and deep network features for the first time, information provided by a single-frame image is better utilized, and the discrimination capability of extracted features is effectively enhanced; the same mode of true and false samples is stripped through the low-rank shared dictionary for the first time, so that the attack detection accuracy is successfully improved, and the method has good generalization.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a face representation attack detection method and system based on fusion features and dictionary learning. Background technique [0002] Today, face recognition technology is widely used in security, payment, entertainment facilities and other scenarios. However, there are certain security risks in the face recognition system. With the development of social networks and the popularization of smart phones, more and more people share personal photos and videos on the Internet. Criminals can use these media to pretend to be other people or deliberately confuse personal identities to attack the face recognition system. To achieve the purpose of infringing on the property safety of others and escaping legal sanctions. Attempts to use legitimate users' photos, videos, etc. to borrow the user's identity through the operation of the face recognition system are called face repr...

AI Technical Summary

Problems solved by technology

Deep convolutional neural networks perform well on image classification tasks, but are limited by the size of live detection data sets. Deep networks supervised only by category labels tend to memorize arbitrary features that exist in the training set, which can easily lead to overfitting. Algorithms poor generalization

Images