Iris Feature Extraction for Personal Authentication Using Macro Features

Abstract

Iris pattern analysis is the most popular technology used in biometric recognition. The main aim of this proposal is to perform the feature extraction of iris by analyzing the macro features such as crypts, radial furrows, lacunae, lesion, etc. Iris has special features which are unique and stable over the individual’s lifetime. Because of this reason, iris pattern analysis has been increasingly used in data security. Iris image classification can be done with the help of Convolutional Neural Network (CNN) through which different patterns of the iris can be mapped. The vector part of the image is analyzed with the help of Capsule networking. The image classification and feature extraction can be done efficiently using CNN and capsule network with the help of Dynamic Routing with Direction and Length (DRDL) algorithm. DRDL is a modified routing algorithm which can be performed by dynamic routing between capsules. The experiments are conducted on PEC database. PEC database consist of about 434 images in which features of iris can be extracted. Formats of the images are in PNG files.