Comparison of Supervised and Unsupervised Learning Methods for Iris Recognition using MATLAB

A stable, authentic, and reliable system is necessary to control and secure areas. This biometric recognition system uses Iris Patterns that differ in every individual for identification. The Iris Recognition system consists of Image Acquisition, Segmentation, Normalization, and Identification. Iris images are downloaded from the CASIA Iris V1.0 database. Image Subtraction and Binarization are done to define the Iris and Pupil boundary. To separate the Iris region from the image, Iris Segmentation is performed by applying Canny Edge Detection and Circular Hough Transform. Daughman’s Rubber Sheet Model converts normalized images into a rectangular block. Here, an unsupervised and supervised learning approach is used for authentication. The entropy of normalized images is utilized for comparison. For supervised learning, we have used the Support Vector Machine classifier to authenticate the image. A Histogram of Oriented Gradients is used to extract the features from the iris patterns and the input image is compared to get results of authentication.