Research On Face Liveness Detection Based On Texture Features And Depth Information

At present,with the continuous introduction and wide application of face recognition related products,the security problem of face recognition system has gradually attracted people's attention.In traditional face recognition system,it is difficult to distinguish the real face and the fake face accurately.Some illegal persons use this vulnerability to spoof attacks on the face authentication system,such as using photos or videos to fake others' identity for profit,which makes the face recognition system a vulnerable object.Therefore,it is very necessary to strengthen the function of liveness detection in face recognition system,which is very important to ensure the safety of the whole system.In order to realize a safe and reliable face liveness detection system,this paper analyzes various methods of face liveness detection at home and abroad,and studies the technology of face liveness detection based on texture features and depth information.The main work and innovation of this paper are as follows:1.Based on the texture features,a method for face liveness detection by combining different color models is proposed,namely the YLH-LBPCM(YCb Cr + Luv +HSV-Local Binary Pattern & Color Moments)method.This method analyzes the differences in texture feature information under different color models.First,the experimental face images are converted into three color models that are useful for distinguishing real faces from deceptive faces.Then the color models are combined to extract features and finally classified.Experimental results show that this method is superior to other literature reference algorithms.2.We have studied the technology of depth data processing and proposed a method of face liveness detection based on depth information fitting plane.Aiming at the imaging difference between the real face and the deceptive attack face,we start from the original depth data and use the method of whether the discrete data point distance value in the depth data can be fitted to a plane to achieve the purpose of liveness detection.In the algorithm design,the performance is improved through group optimization.Experiments show that the method has high accuracy,simple calculation and good portability.At the same time,a face liveness detection network based on depth information extraction feature is constructed by using convolution neural network training method.By visualizing the original depth data into gray images,the powerful function of convolution neural network is used to automatically extract features for classification and recognition.Experiments show that the network is competitive compared with the reference method.Finally,the two methods are analyzed and compared.3.We designed a face liveness detection system based on texture feature and depth information fusion.By analyzing the advantages and disadvantages of texture features and depth information,and merging the advantages of the two technologies,a multi-modal living body detection system was designed based on kinect camera,and the detection and prevention capabilities were enhanced.The face liveness detection method proposed in this paper can be embedded in the face recognition system module,which can effectively enhance the reliability and security of the face recognition system.At the same time,it is also of great significance for studying multi-modal face liveness detection.