A method and system for rapid living detection

Abstract

The invention discloses a method and a system for rapid living body detection, belonging to the field of living body detection. By constructing the spectrogram of the real face image and the multi-scale and multi-rotation angle of the second-shot face image, and using the projection vectors of the spectrogram in the horizontal and vertical directions as the feature vector, the moiré pattern of the image can be fully extracted. Features, without relying on complex neural network structure, to achieve rapid detection of video replay attack methods. The blood volume distribution is constructed by using the correlation between the overall facial signal of the face image and the local signal of interest. On this basis, the frequency peak distribution is calculated. Combined with the threshold segmentation model, the precise positioning of the area containing the physiological signal is realized, which can be used for high simulation. High-precision detection of mask and high-definition image attacks. The present invention does not need to specify facial actions to cooperate with detection, and can perform comprehensive detection only by collecting facial video, covering the main living body detection attack methods, and has high detection accuracy and detection speed.

Description

technical field [0001] The present invention relates to the field of image processing, and in particular, to a method and system for rapid living detection. Background technique [0002] At present, the realization of living attack detection based on signal processing technology and deep learning technology has high practical value, so it is a hot and difficult point of current research. Living body detection is mostly used in security verification, which has high requirements on the detection accuracy, detection speed and detection range of living body attacks. It often requires the object to be detected to randomly perform specified facial movements to cooperate with the detection, but this detection method cannot completely prevent high Accurate simulation mask attacks and other means, and users are prone to discomfort or embarrassment during the detection process. A common problem in the prior art is that the detection method or means is single, which cannot effectively...

AI Technical Summary

Problems solved by technology

Liveness detection is mostly used in security verification. It has high requirements for the detection accuracy, detection speed and detection range of live attacks. It often requires the subject to be detected to perform random specified facial movements to cooperate with the detection. However, this detection method cannot completely prevent high Accurate simulation of mask attacks and other methods, and users are likely to feel uncomfortable or embarrassed during the detection process

Common problems in the existing technology: the single detection method or method cannot effectively cover most of the typical living detection attack methods, and the single detection method is prone to obvious loopholes, which are easy to be identified and exploited

Images