An Intelligent Iris Recognition Technique

Biometrics are vital in security. Facial recognition, fingerprints, and iris recognition are all examples of computer vision biometrics. Unique authentication based on iris structure is one of the finest approaches for iris identification. This research provides an iris-based biometric identification system combining CNN and Softmax classifier. The system consists of picture augmentation by histogram equalization, image reduction by discrete wavelet transformation (DWT), segmentation by circular Hough transform and canny edge detector, and normalizing by Daugman's rubber-sheet model. Each picture is adjusted before being fed into the DenseNet201 model. The Softmax classifier then sorts the 224 HID iris classes into 249 CASIA-Iris-Interval classes, 241 UBIRIS.v1 iris classes, and 898 CASIA-Iris-Thousand classes. The performance of our suggested system is determined by the setting of its deep networks and optimizers. In terms of accuracy, it exceeds existing approaches by 99%.