Two-dimensional face fraud detection classifier training and face fraud detection method

Abstract

The present invention provides a method for generating a two-dimensional human face fraud detection model. The method includes: first, preprocessing all human face pictures in the training set to obtain normalized human face images; Extract LBP eigenvectors, Gabor wavelet eigenvectors and one-dimensional pixel eigenvectors from the integrated face image; thirdly, splice these three eigenvectors to form the final eigenvector; fourthly, use support vector machine to The final feature vector is trained to obtain a two-dimensional face fraud detection classifier; this method extracts the feature information of the difference between the face and the photo; the feature extraction is simple and efficient, does not require the user's deliberate cooperation, and can be used in low-resolution situations Get good results. In addition, based on the two-dimensional face fraud detection classifier obtained by the above method, the present invention also proposes a face fraud detection method, which has the advantage of high detection accuracy and can effectively prevent face fraud.

Description

technical field [0001] The invention relates to the fields of computer vision and graphic image processing, in particular to the training of a two-dimensional face fraud detection classifier and a face fraud detection method. Background technique [0002] At present, two-dimensional biometric technology (ie, recognition based on two-dimensional face biometrics) is a very important research field. Perspective change, occlusion, and complex outdoor light have always been the difficulties of face recognition. Although a lot of work has been done to solve these problems, the fraud attack vulnerability of the face recognition system has been ignored by most systems. Face recognition systems rely on flat graphics for identity detection, and the system is vulnerable to fraudulent attacks from printed or electronic photos. For example, Windows XP and Vista laptops from Lenovo, Asus and Toshiba all have built-in web cameras and biometric systems that authenticate users by scanning ...

AI Technical Summary

Problems solved by technology

Although there are more and more researches in this field and some public and open databases have been released one after another, there are not many standard databases that can provide relatively objective development tests for fraud detection algorithms. There is no consensus on fraud detection algorithms

The current research methods for face fraud detection are either too complicated to be practical (requires real-time and fast processing in practice), or use some unconventional imaging systems (multispectral imaging) and high-resolution cameras, which are not available. Conditions for practical application

Images