Robust iris recognition using deep learning and SVM classification

The main issues iris biometric identification systems suffer from are accuracy and reliability. The proposed system is considered to find the solution for these two concerns to increase accuracy and improve reliability and performance using the enhanced Iris biometric. This study presents the design and implementation of a biometric recognition system based on iris features to recognize persons using the Dughman model to localise the iris region in the pre-processing stage CNN algorithm to feature extraction and SVM to classification. Initially, the pre-processing phase of this recognition system includes iris localisation, segmentation, normalisation, and feature enhancement by using histogram equalization. Then the CNN architecture effectively extracts features to be classified later by the SVM. The system's performance was evaluated utilizin g the Ahmed Myiesr Fathi (AMF) iris database. The proposed system achieved outperformance compared to state-of-the-art iris recognition models. Notably, the system achieved 96.2% and 97.93% identification rates for non-enhanced images and enhanced images, respectively. The Equal Error Rate (EER) for non-enhanced images is 1.03, while for enhanced images, it is nearly zero. These results show a substantial improvement in recognition performance resulting from feature enhancement in the pre-processing phase. In addition to use deep learning and machine learning algorithms advanced.