Research On Face Anti-spoofing Algorithm Based On Video

With the development of technology,biometrics has been widely used in various authentication fields.Face,as a non-invasive and highly recognizable biometrics,has also been applied to various authentication occasions.Face recognition has also become a biometric means under various occasions.However,there are problems in the security of face recognition systems.Face information can be detected by individual photos and video before face recognition system to increase the ability of face recognition.Therefore,it is necessary to add the steps of in face anti-spoofing before face recognition to increase the anti-deception ability of face recognition and ensure the safety of face recognition.1.A face anti-spoofing algorithm using binocular camera ranging is proposed,which consists of two parts.The first part combines facial landmark detection with blink detection using frame difference method to solve the problem of false detection in blink detection.Effectively distinguishes photo visits.The second part,aiming at the problem that blink detection can not distinguish video,uses a binocular camera to measure distance,performs image distortion correction and performs pixel matching through the human face landmark detection method to obtain the distance of landmark,effectively distinguishes between real visits and video forged visits displayed on the screen.2.A method for face detection in face of general cameras is studied.The existing face database has a small amount of data and the lighting conditions are quite different.This paper has established a face anti-spoofing detection database including photo spoofing and video fraud to reduce environmental impact.Most face detection methods use manual texture features for classification,but the depth features are more robust to this complex classification problem.Combining texture methods with CNN networks,LBP texture descriptors are integrated into the modified in the convolutional neural network,deep texture features are extracted from the image,and good performance is obtained on various human face detection databases.3.Aiming at the problem of the general camera under the equipment and the lighting environment is complicated,a living face detection method that requires additional near-infrared camera device is proposed.The general camera and the near-infrared camera are used together.In the LAB color space,the histogram color feature of the face under the general camera is extracted,and the LBP texture feature of the face is extracted under the near-infrared camera,and the final discrimination of the color and the texture is performed using the support vector machine,achieving an accuracy and practicality.Very high living face detection system.