Research On Face Liveness Detection Based On Visible Light Imaging

With the rapid development of face recognition technology,face recognition has been gradually applied to security,transportation,finance and other fields.Face recognition technology brings great convenience to our daily life.At the same time,its security is getting more and more attention.Currently,face recognition technology is mainly based on image information,which makes the technology vulnerable to attacks such as photo attacks,video attacks and mask attacks.Face liveness detection analyzes the face image to determine whether the subject is a living body,which is an important technical method to protect face recognition technology from attack and it has been an important part of the face identity authentication system.This thesis summarizes the research results of human face liveness detection.In view of the fraud problem in face liveness detection,this thesis studies the features of face texture,anisotropic diffusion and deep learning,and designs two kinds of face liveness detection algorithms.The main work of the thesis is as follows:1.A face liveness detection algorithm based on image anisotropic diffusion feature and texture information is proposed,which mainly includes image preprocessing module and feature extraction module.The real face is a three-dimensional structure,and the fake face image is a two-dimensional plane.The light variations on the surface are different,and anisotropic diffusion can extract this feature well.In this thesis,an anisotropic diffusion is used to construct an image diffusion velocity model,and the LBP algorithm is used to extract the diffusion velocity characteristics of the image,and the support vector machine is trained.At the same time,the thesis also extracts the blur degree feature and color texture feature of the face image,fuses the two features by the method of feature matrix cascading,and trains another classifier.Finally,a decision is made based on the result of the classifier output probability weighted fusion.The half error of our method on the two public datasets is 5.76% and 3.54%,respectively.2.A face liveness detection algorithm based on improved deep learning model and texture features is studied,including deep learning feature extraction and image texture feature extraction.Different from the previous face liveness detection algorithm,this thesis believes that the background area of the image also contributes to face liveness detection.In order to make better use of image features at different depths in convolutional neural networks,a bypass structure is added to the convolutional neural network,and the improved neural network is used to extract the depth features of the image and train the classifier model.The Local Phase Quantization features of the image have great advantages across the data set.Therefore,this thesis extracts the Local Phase Quantization features of the image and trains the SVM classifier.Finally,a decision is made based on the result of the classifier output probability weighted fusion.The experimental results show that our method can effectively detect fake face images,which further improves the generalization ability of the face detection algorithm.