Two-stage anti-cheat detection method based on multi-angle face and environmental information

Abstract

The invention discloses a two-stage anti-cheat detection method based on a multi-angle face and environmental information, comprising the following steps: tracking a face in a detection area, detecting the face in multiple angles, and classifying, marking and storing the detected face images according to angle; determining a background contrast area according to a detected face location; contrasting the background of a current face and the background under an unmanned environment in the background contrast area; judging whether the detected face satisfies a multi-angle face condition, and judging whether the matching degree between the background of the current face and the background under the unmanned environment in the background contrast area satisfies a condition; judging that the current detected face is a real face when both conditions are satisfied, or, judging that the current detected face is a face picture or video; and storing a front image of the detected multi-angle face. The method is more user-friendly in operation, can be used to perform face anti-cheat detection quickly and efficiently, and guarantees the safety and reliability of a face recognition system.

Description

technical field [0001] The invention relates to the technical field of face recognition, in particular to a two-stage anti-spoofing detection method based on multi-angle faces and environmental information. Background technique [0002] Among all biometric identification technologies, face recognition is the most natural and visual biological identification method, which is in line with human's own physiological and visual habits. Humans themselves identify people through their faces. In recent years, face recognition technology has developed rapidly and has been more and more widely used, and the security of face recognition has also received more and more attention. Spoofing attacks are a fatal threat to face recognition systems. After decades of development, in the field of face recognition, scholars have proposed a large number of live detection technologies and methods to be applied to enterprise attendance face recognition systems with small user capacity. Places, suc...

AI Technical Summary

Problems solved by technology

Applying this system for checking attendance will greatly increase the cost of equipment, and it will take a long time to detect each person's living body, and it is strongly dependent on the cooperation of users, which is not conducive to the application in places with large user capacity and high traffic; the authorized announcement number is CN 100514353 C "A Liveness Detection Method and System Based on Human Face Physiological Movement" proposed a living body detection system, which distinguishes real faces from photo faces by detecting the physiological activities of faces in the detection area. And if the video spoofing attack is carried out to this system, its living body detection method can be cracked, making the system unreliable; It needs to be trained separately for specific people, that is, each registrant corresponds to a live face classifier. This method is also only suitable for enterprise attendance with a small user capacity or special department attendance. When the number of users is too large, it is obviously for It is very uneconomical for each individual to train a live classifier

Images