Retinal biometric identification using convolutional neural network

Authentication is needed to enhance and protect the system from vulnerabilities or weaknesses of system. There are still many in use traditional authentication methods such as PINs passwords, being hacked. New biometrics used deal with this problem. Biometric characteristics using retinal identification unique difficult manipulate compared other biometric iris fingerprints because they located behind human eye thus reach by normal vision. This study uses fundus image blood vessels that have been segmented for its features. The dataset sourced DRIVE dataset. preprocessing stage extract features produce an vessel segmentation. resulting segmentation carried out a two-dimensional transformation process rotation, enlargement, shifting, cutting, reversing increase quantity sample image. results resulted 189 images divided details ratio 80 % 151 training data 20 38 validation data. forming research model Convolutional Neural Network method. built during consists 10 iterations produces accuracy value 98 %. model's identifying individual retinas