| Iris Recognition System for Biometric Identification: A Gabor Wavelet and Support Vector Machine Approach | |
|
MSRDG International Journal of Computer Scientific Technology & Electronics Engineering
© 2025 by MSRDG IJCSTEE Journal Volume 1 Issue 3
Year of Publication: 2025 |
Paper Download Article ID MSRDG-IJCSTEE-V1I3P104 |
|
Abstract: Biometric identification has become indispensable in high-security environments, and the iris—owing to its unique, stable texture—is widely regarded as one of the most reliable biometric identifiers. Despite decades of research, many deployed systems continue to struggle with accuracy under real-world conditions involving occlusion, variable illumination, and non-ideal subject cooperation. This paper presents an end-to-end iris recognition framework integrating Daugman's integro-differential operator for precise boundary localisation, rubber-sheet normalisation for pose invariance, multi-scale Gabor wavelet decomposition for texture encoding, and a multi-class Support Vector Machine (SVM) classifier for identity verification. The proposed system was evaluated on three publicly available benchmark datasets: CASIA-IrisV4, UBIRIS v2, and MMU2. Experimental results demonstrate a Genuine Acceptance Rate (GAR) of 99.12%, False Acceptance Rate (FAR) of 0.08%, and False Rejection Rate (FRR) of 0.79% on CASIA-IrisV4. The system achieves a mean recognition accuracy of 99.12% across subjects and a total end-to-end processing latency of 115 ms on commodity hardware, outperforming existing state-of-the-art baselines including CNN-based and Local Binary Pattern (LBP)-based methods. The proposed hybrid framework establishes a new performance benchmark for near-infrared iris recognition and is well-suited for deployment in border control, access management, and forensic identification scenarios. |
|
| Keywords: Iris recognition · Biometric identification · Gabor wavelets · IrisCode · Daugman operator · Support vector machine · CASIA-IrisV4 | |
