Retinal biometric system using electromagnetism-like optimization algorithm

Abstract

Biometric system is the technology used for the purpose of identifying the physiological and behavioural characteristics of an individual as input, analyzes it and identifies the individual as a genuine or imposter. Among all biometrics, retina based identification is perceived as a robust, unforgeable and reliable form of biometric solution. The blood vasculatures of retina are unique and used as features for retinal biometric system. In this work, an attempt has been made to employ an Electromagnetism-like Optimization Algorithm (EMOA) with Otsu Multilevel Thresholding (MLT) for segmentation of vascular pattern from the retinal fundus images for retinal biometric system. Retinal images are taken from the publicly available database such as DRIVE, STARE and HRF. The original images are subjected to preprocessing. Segmentation is carried out on the preprocessed images using EMOA Based Otsu MLT. This method provides comparatively better segmentation accuracy of 0.974 than other existing methods. Texture and vessel features are extracted from the segmented image. Matching is done between query and enrolled images using Euclidian distance measure. Decision is made using best matched image. This biometric system shows matching accuracy of 97%. Hence, this method could be recommended for retinal biometric system.