EnhanceDeepIris Model for Iris Recognition Applications

Abstract

In this study, an iris recognition technique based on enhanced EnhanceDeepIris model is proposed in order to examine iris recognition at a deeper level. The process first uses convolutional features to extract human iris features. Then, in order to solve the deformation problem caused by the radial data of iris texture, the sequential metric model is introduced to realize the effective recognition of iris. Finally, the structure of the model is thoroughly examined to investigate its actual performance. The study found that after running the four algorithms on the ND-IRIS-0405 and CASIA-Lamp datasets, the loss function value of the research method began to approach its lowest value at the 14th and 9th iterations, while the other algorithms continued to slowly decrease. Additionally, at the 5th and 4th iterations respectively, the accuracy of the research method was nearly 91.00%.After 20 classification predictions, the average recognition accuracies of the research method proposed in this paper, iris segmentation method based on end-to-end multi-task segmentation network IrisST-Net, iris recognition method based on full complex-valued neural network, and iris recognition method with hybrid preprocessing and feature extraction were 99.88%, 98.72%, 97.47%, and 89.77%, respectively. These findings suggest that the study approach recognizes the iris the best and can identify its primary contour information with accuracy. The results of the application demonstrate that the research method provides the clearest detection of the iris edge, with less noise, and can accurately detect the main contour information of the iris. This provides a reference for optimizing related technologies in the field of image recognition.

Keywords

• EnhanceDeepIris model
• deep learning
• iris recognition
• feature extraction