Fuzzy-CNN: Improving personal human identification based on IRIS recognition using LBP features

The iris functions as a resilient instrument for dependable human identification, showcasing substantial promise in recognizing individuals with a considerable level of assurance. The crucial step in iris recognition lies in extracting effective features. Traditionally, various handcrafted features, devised by biometrics specialists, have been employed for implementing iris recognition systems. However, given the remarkable success of Fuzzy-deep-learning in addressing computer vision challenges, local binary patterns (LBP) features learned by Convolutional Neural Networks (CNNs) have garnered considerable interest for application in iris recognition systems. This study evaluates the LBP features followed by the Fuzzy-CNN model for classification. The system's performance is compared with several machine and deep learning models. The proposed model obtained an accuracy of 99.55%, 98.85% precision, 99.47% recall, and 99.22% F1 Score. Rigorous testing is conducted on four public datasets, namely IITD, CASIA-Iris-V1, CASIA-Iris-thousand, and CASIA-Iris-V3 Interval. The proposed iris recognition system demonstrates outstanding results, achieving a notably high accuracy rate.