A video human face living body detection method

Abstract

The invention provides a video human face living body detection method. The method comprises the steps of inputting a video stream; cutting the video stream to obtain N video frame images; establishing N-1 dynamic models for the N video frame images by using a dynamic correlation model algorithm; performing model selection based on phase angle to obtain a final single dynamic reduced-order model image; extracting an LBP uniform pattern feature histogram based on segmentation weighting from the dynamic reduced-order model image; discriminating virtual attack from a legal user based on the SVM classification of three kernel functions. Compared with the prior art, the method uses the dynamic correlation model to pre-treat video frames for the first time, can accurately capture dynamic changes of the human face and better makes up for the blank of lack of living body detection methods with better performance for video attacks in the prior art; the method gives prominence to the human face area of great importance for the living body detection performance. The human face living body detection performance algorithm is generally superior to other conventional methods for video attacks.

Description

technical field [0001] The invention belongs to the field of image processing, and relates to a living body detection technology in face recognition, in particular to a method for resisting face counterfeiting video attacks by using face dynamic information and micro-texture features. Background technique [0002] In recent years, face recognition technology has developed by leaps and bounds, and it has higher stability in response to changes in posture, illumination and expression, which has also prompted more and more occasions to use face recognition technology for identity authentication. [0003] However, the existing face recognition systems still lack reliable security, because it is extremely vulnerable to various forms of false attacks by illegal users, the most common ones are face photo attacks, face video attacks, and three-dimensional face recognition systems. Model attack. Therefore, research on the combination of reliable face detection technology and face re...

AI Technical Summary

Problems solved by technology

[0008] For above-mentioned prior art and existing problem, the present invention proposes a kind of video human face living body detection method, combines dynamic correlation model and LBP (LocalBinaryPattern) equivalent pattern (UniformPattern) feature extraction, feature vector is sent into support at last The vector machine SVM (SupportVectorMachine) classifier training and testing scheme solves the shortcomings of the current existing methods, such as unstable detection, easy to be affected by the external environment, and weak resistance to video attacks, and its detection performance is better than the current existing methods

Images