A Study on Recognition of

Biometric research has experienced significant advances in recent years given the needfor more stringent security requirements. Iris recognition has been demonstrated tobe an efficient and reliable technology for personal identification. In this thesis weemployed three new matching schemes for iris recognition, the Scalar Product (SP),the Multi-dimensional Artificial Neural Networks (MDANN), and the Elastic GraphMatching (EGM).These three methods are trained and tested using two databases of gray scale eyeimages (CASIA and UBIRIS). They are trained using 996 and 723 iris images fromthe CASIA and UBIRIS database respectively. We have tested them using 915 and448 iris images from the CASIA and UBIRIS database respectively.We have found that, there are 81 and 34 iris images from the CASIA and UBIRISdatabase respectively, are not used at all because of the failure analysis of locatingiris for different causes. The Correct Recognition Rate (CCR) for the SP matchingmethod is 98.26%, the CCR for the MDANN is 99.25%, and that for the EGM is98.79%.