Anti-spoofing Technology Of Iris Recognition

With the global attention and top-level layout of biometric identification technology,the research and application of biometric identification are developing rapidly and gradually replacing the traditional identification methods.Iris recognition is one of the most important biometric recognition,which is widely used in attendance,security,access control,identity detection and other aspects.However,it is an inevitable risk and challenge for the development and application of iris recognition technology that how to detect and deal with the deceptive attack of physically synthesized printed iris,iris-glasses and others on iris recognition system.Based on the above background,this thesis proposes specific anti-deception methods for printed iris attack and iris-glasses attack,which enriches and improves the anti-deception methods for iris recognition:1.After collecting and sorting,the self-use experimental database was established,including PRIP database used for research on anti-spoofing of printed iris attack,and DLRIP database used for research on iris anti-spoofing of glasses iris attack.2.Aiming at the printed iris attack,an iris texture analysis algorithm based on deep convolutional neural network is proposed.A depth CNN model for iris texture analysis is constructed and the training process is visualized.Experimental analysis shows that the model has short training time and fast convergence speed,and the recognition rate of 3200 real iris and 3200 printed iris in PRIP database reaches 99.54%.Compared with the traditional iris texture analysis algorithm,the recognition rate is improved.Compared with other models,the learning speed and convergence speed are faster.3.Proposed a anti-spoofing method of glasses iris attack based on depth information,including the anti-spoofing of glasses attack and the anti-spoofing of lenses attack:1)Aiming at the frame glasses iris attack,the dispersion characteristics of frame glasses samples which are different from other samples is analyzed,proposes an glasses attack anti-spoofing algorithm based on mean square error statistics.2)Aiming at the contact lens attack,the dispersion characteristics of contact lens samples which are different from other samples is edge depth transient characteristics,proposes an contact lens attack anti-spoofing algorithm based on pass-mean detection.Experimental analysis shows that the identification rate of the above method for frame glasses samples and contact lens samples in the DLRIP database reaches 100%,reduces the threat of glasses attack to iris recognition system.