Personal identification based on fusion of iris and periocular information using convolutional neural networks

Iris recognition, a pivotal component in computer vision, serves as a vital tool for verification and identification in security and forensic science domains. Despite itswidespread application, current iris recognition systems face significant performance limitations, especially under adverse conditions like suboptimal lighting, partial iris visibility, eye misalignment, and scenarios involving greater distances. These challenges often impede the effectiveness of traditional iris recognition methods. However, the emergence of advanced neural network-based image recognition technologies, particularly convolutional neural networks (CNNs), has opened new avenues for addressing these limitations. CNNs excel in extracting intricate features from images, thereby enhancing recognition performance to levels that either match or exceed those of conventional iris recognition techniques. In our research, we introduce an innovative methodology that integrates traditionaliris recognition with periocular information through the application of CNNs. This fusion aims to overcome the aforementioned challenges effectively. To validate our approach, we employed three distinct datasets: CASIA-Iris-Thousand, CASIA-Iris-Lamp, and UBIPr. The results from these evaluations demonstrate a significant improvement in accuracy and reliability over standalone iris recognition systems. Furthermore, our paper delves into the practical implications of deploying such an advanced system in real-world scenarios. We explore the potential applications and benefits of this enhanced iris recognition technology, underscoring its suitability for highsecurity environments where accuracy and reliability are paramount. This research not only contributes to the field of computer vision but also sets a new benchmark for future developments in biometric security systems.