Liveness Detection For Iris Recognition Technology

With the fast development of artificial intelligence and image identification technique and the vigorous growth in the penetration rate of intelligent electronic devices,advanced biometrics are gradually merging into everyday life.In the recent 30 years of fast development of the iris recognition technology,several classical algorithms are proposed and improved by experts across the world.However,no matter how advanced the biometrics are,they cannot be considered practical if they fail to detect and react to forgery attacks.Thus,this work tackles the limitations of the traditional biopsy algorithm based on the pupil's reflection of light by introducing the classifying features of the iris texture structure in respond to the corresponding attack model.For the intelligent mobile electronic devices that the traditional method fails to protect effectively,a new liveness detection method based on the double-band near infrared band method is proposed,which improves the protection against various attacking methods.The novelties of this work are as follows:(1)This work proposes a new liveness eye detection method based on the iris texture.This work combines the traditional biopsy algorithm based on pupil's reflection with the recognition of the iris texture to defend attacking methods based on mimicking the pupil contraction rate.An adaptive threshold method is applied to analyze pupil which yields better robustness comparing to the fixed threshold method and ensures the accuracy of following identification processes.This work also proposes an advanced Hough transform method algorithm combine with the morphological centroid,which only locate the iris and pupil in the region of interest.This lowers the dimension of the searching and achieves higher efficiency and accuracy comparing to traditional geometry algorithms.(2)This work proposes a liveness iris detection method based on the double-band near infrared band which can distinguish artificial samples from the real blood vessel textures by examining their absorption and reflection rate to near infrared light with different wavebands.By experimentally identifying the two near infrared wave bands where the largest difference occurs for the texture of real blood vessels and artificial samples,the texture of the blood vessels is well characterized.The method can effectively protect intelligent electronic devices from new attacking methods which defeat traditional protections.In this work,the approaches proposed above are studied in detail using data collected experimentally and data from the CASIA v1.0 and v2.0 iris database,which validates the robustness of these novel approaches.