Retinal Blood Vascular Pattern Based Personal Identification (T-0679) (MFN 3604)

Abstract

Digital image processing, the immense field of engineering provides wide applications in Biometrics Engineering and Security. Biometrics technologies are mostly used for the certification and attesting purposes for secure access control. In early ages, biometrics was considered as flawless in comparison with keys, passwords, codes or pin numbers security measures, with least significance of imitation and deception in them. Due to physical exposure of the regions like fingerprints, iris or ear geometry recognitions can get burned and damaged. The rare and unique security system is addressed in this research, as the retinal vascular patterns for personal identification is a reliable technique. In this research, we present a methodology for retinal recognition on bases of blood vessels which has recently become popular for fraud detection and prevention, The proposed methodology is divided into two divisions i.e. enrollment and identification. The enrollment phase includes addition of new person in database whereas in identification phase, test fundus image is recognized by matching it with entries in databases. We propose a novel system for analysis of fundus images to represent it with reliable feature sets. The proposed system first enhances and segments vascular pattern from input retinal image using Gabor wavelets and multilayered thresholding technique. The feature points i.e. vessel endings and bifurcation are extracted and validated using crossing number method and a novel windowing based method respectively. We also propose a technique to filter false structure such as short vessels, breakages and spurs to improve the matching score. One main problem with retina recognition is that only one database (VARIA) is publicly available for retina recognition, so we have developed our own database with the help of AFIO (Armed Forces Institute of Ophthalmology) and named it as RIDB (Retina Identification DataBase). The proposed system is evaluated using these databases and results show the significance of proposed system.