Personal Identification Based on Fusion of Iris and Periocular Information Using Convolutional Neural Networks

Abstract

Iris recognition has long been a security verification and identification tool in the field of computer vision, frequently utilized in security and forensic science applications. However, the performance of current iris recognition systems remains unsatisfactory, particularly when capturing images under challenging conditions such as poor lighting, iris occlusion, eye misalignment, or increased distance. Despite this, recent advancements in neural network-based image recognition systems, such as convolutional neural networks (CNN), have shown the potential to tackle these problems by extracting complex features within images and delivering performance that is either comparable to or surpasses that of traditional iris recognition methods. In this paper, we proposes a novel approach that fuses iris recognition techniques with periocular information using CNN to address these challenges. Our approach was evaluated by using the CASIA-Iris-Thousand dataset. The evaluation results showed that the accuracy was improved from 93.64% to 99.80% compared with using the iris recognition alone. The false acceptance rate and false rejection rate were reduced noticeably from 6.30% to 0.17% and from 12.33% to 5.00%, respectively. We also discuss the potential for real-world application of such a system.