Iris Recognition System (IRS)

In response to the growing demand for secure and reliable biometric authentication, this paper explores and compares various iris detection techniques that form the foundation of modern iris recognition systems. Owing to its uniqueness and stability over time, the human iris has become a trusted biometric trait for identity verification. This study systematically examines a spectrum of detection methods, from classical image processing techniques such as the Hough Circle Transform and Daugman’s operator, to machine learning models utilizing handcrafted features, and modern deep learning architectures like U-Net, YOLOv5, and DeepLabV3+. Leveraging standard datasets including CASIA-IrisV4, UBIRIS.v2, and IITD, the research evaluates each method based on detection accuracy, computational efficiency, and robustness in challenging scenarios, including occlusions, poor lighting, and motion blur. Results show that deep learning models not only achieve superior accuracy but also exhibit faster processing and better generalization, making them highly suitable for real-time and mobile biometric applications. Furthermore, the study highlights the benefits of combining traditional and deep learning approaches to create hybrid models that balance speed, accuracy, and resource usage. The analysis also emphasizes the importance of data augmentation and preprocessing in enhancing model performance across diverse imaging conditions. These insights pave the way for developing scalable and user-friendly iris recognition systems adaptable to real-world deployments.