CNN-Based Feature Extraction and Classificati...

CNN-Based Feature Extraction and Classification for Periocular Biometric Recognition

Author Name : Mrs. Y. Karuna Manjusha, V. Jagadish Sri Siva Sai, T. Jagadeeshwar Reddy, G. Sri Datta Sai, R. Surendra Babu, M. Vamsi

ABSTRACT The world has been affected by the corona virus epidemic, which soon turned into a pandemic. The need for face masks and social isolation highlight the need for contact free biometric authentication for all future systems. Periocular biometric, which does not require physical contact is a solution in such scenarios because it can detect people wearing face masks. The term "periocular region" defines the area surrounding the eye comprising several components including the sclera, eyelids, lashes, brows, and skin. Our method involves extraction of the region of interest by cropping the periocular region of every facial image. CLAHE is then performed on each periocular image extracted. VGG16 and ResNet50 CNN models are used for feature extraction. The picture is classified using KNN classifiers. Our experiments demonstrate that the proposed approach achieves an accuracy of 80% with ResNet50 and 75% with VGG16onachallenging dataset consisting of 24 subjects with significant variations in pose, illumination, and expression. The suggested approach offers a promising solution for noninvasive and contactless biometric identification, which can have broad applications in security and surveillance systems.