IRIS BASED BIOMETRIC SYSTEM FOR HUMAN IDENTIFICATION

Abstract

There is an increase in security concerns in security concerns on issues such as identity theft indicates the needs of a new reliable security system. This can be accomplished by using biological features of Human via Biometrics such as Iris recognition. In the first place, these thesis review literature on Iris scan Technology comparing it with other image based biometrics. The first aim of implementing this project is to reconstruct an iris recognition system for Near Infrared (NIR) image based on the Daugman’s algorithm. After finishing this, the second aim is to extend the finished system to handle color spectrum image and improved on individual sections. Daugman’s iris recognition system is based on the location of the iris boundaries via a circular edge detector utilizing integrodifferential operators. The iris is then normalized using a rubber sheet model. After projecting onto Gabor filters, phase information is extracted and stored in a bit code. Hamming distances between the input image code and stored codes are calculated. Based on a test of statistical independence the irises are matched or not. Noisy areas of the iris are masked off from analysis via mask bits. Combining these elements an iris recognition system was implemented. The system produced in this project was written in Matlab. The database used in this project is from CASIA and UBIRIS, only around 70-80% can locate the iris correctly by the program developed in this project. Out of those images, about 75% of them can be recognized correctly without accepting any false matches. Therefore, the iris recognition system developed in this project was partly successfully. The failure to recognize all color images is because of the poor localization and designing the Gabor filter parameters using a small database. However, the system has a good processing time on comparing the irises.