Open AccessComputational Methodology for Iris Segmentation and Detection in Images from the Eyes Region Using Convolutional Neural Networks
Author(s) -
Fredson Costa Rodrigues,
Anselmo Cardoso de Paiva,
João Dalysson S. de Almeida,
Geraldo Bráz,
Aristófanes Corrêa,
André Soares
Publication year - 2021
Language(s) - English
Resource type - Conference proceedings
DOI - 10.5753/wvc.2021.18910
Subject(s) - iris (biosensor) , iris recognition , artificial intelligence , computer science , computer vision , biometrics , convolutional neural network , segmentation , pupil , eye tracking , pattern recognition (psychology) , neuroscience , biology
Eye tracking is an application of computer vision responsible for detecting the iris and pupil in the eye region. The usefulness of this tracking contributes to research that assesses cognitive aspects through pupillary reactions identified in these detected regions. Another application in this task is iris recognition in digital biometrics. This study aims to carry out the verification and detection of the iris in images of the eye region occluded by eyelashes, eyelids and specular reflexes, through a deep neural network called At-Unet in this article. In order to assist in eye tracking this method achieves 95.32 % of data coefficient when segmenting the iris of the eyes, indicating the efficiency of this methodology.