Research On Methods Of Liveness Detection In Face Authentication System

With the development of microelectronics and the deeper research of visual system,the stability and efficiency of face authentication technology has been greatly improved.As a result,face authentication technology has been widely used in security,finance,network security,property management and attendance and other fields.But at the same time,the means of attacking face authentication system are endless.For instance,the photo or the video of the legitimate person can be used to pose a threat to the face recognition system.If the face authentication system is lack of effective liveness detection,the attacker can only rely on photos and video to cheat the trust of the system,which will bring immeasurable loss of legitimate users.Therefore,the study on the liveness detection of face authentication system to distinguish the real face from photos or video has a very important significance.By analyzing the recapture characteristics of photos or videos,this paper analyzes the difference of real user imaging and photo or video imaging,and designs an efficient and reliable liveness detection algorithm from the aspects of pretreatment,feature extraction and multi-feature fusion.The main contents are as follows:1.As the complex background and illumination condition of face authentication system in practical,this paper adopt a method which has robustness to complex background and illumination to preprocess images captured by face authentication system,distinguish the face part from the background part and extract the face image.Then,normalize the extracted face part.2.According to differences in the texture between recaptured face images through the face detection system and liveness face images collected directly,in this paper,multi-scale dynamic texture analysis is used to extract the dynamic texture information and the difference between the real face and the photo or video in the feature space are compared in this paper to maximize the feature difference.The experimental results show that the proposed method can distinguish the real face from the image and video face effectively.3.In order to further improve the accuracy and efficiency of detection and the generalization ability of the method,a multi-feature fusion method based on motion information and texture information is proposed in this paper.Non-rigid motion analysis extracts facial motion information.Facial consistency analysis of the background is obtained from facial movement and background movement correlation.Multi-scale analysis represents the dynamic texture of the image difference.Because each feature clearly has its own meaning,the method has a stronger generalization ability.Experiments show that the proposed method achieves better detection accuracy by combining multiple features.