Iris Recognition using Hough Transform and Neural Architecture Search Network

Biometric authentication technology is a key in security devices and systems. Each individual has unique biometric identifiers and irises are among them. Each Iris is unique in its own. In order to exploit them, for the purpose of identification, precision plays an important role to avoid errors. In this paper, a model has been proposed for recognizing irises and is capable of distinguishing and filtering only relevant images. These images are processed through steps of segmentation and classification. Through the combined means of Deep Learning Neural Architecture Search Network (NASNet) and Morphological Image Feature Extraction techniques, the developed system performs real-time iris segmentation and detection with cent percent accuracy. The process begins by taking input image and running through the segmentation process, where the Region of Interest (RoI) Iris images are extracted from the CASIA Iris Interval Dataset and the output is fed to the applied Convolutional Neural Network (CNN) for classification and identifying the batch of images. The Iris Recognition model has been validated and trained successfully.