Developing Iris Segmentation By Using U-Net Model

Abstract

Biometric Authentication Stands Out As A Superior Method Compared To Traditional Identification techniques Due To The Inherent Uniqueness Of Various Physiological And Behavioral Traits In individuals Additionally, Biometric Markers Such As Fingerprints, Facial Patterns, Hand Geometry,Retinal Scans, Palm Prints, Voice, Gait, Signatures, And Keystroke Dynamics Are Commonly Employed for Authentication. On The Other Hand, Conventional Security Measures Like Token-Based Systems, Encoded Cards,Passwords, And Pins Suffer From Numerous Limitations. Hence, The Primary Goal Of This Study Is Toemploy Deep Learning Techniques For Iris Segmentation, Aiming To Enhance Image Quality And Overall biometric System Performance. The Study's Scope Is Specifically Focused On Iris Segmentation Using machine Learning Models Due To The Complexity Of Iris Recognition Systems. The Proposed Iris Segmentation System Comprises Several Key Components Including Imageacquisition, Iris Segmentation, Feature Extraction, And Matching For Recognition. Datasets From repositories Like CASIA Were Utilized, Supplemented By Locally Collected Images From Volunteers for Testing Purposes. Limitations In Image Quality Due To Factors Like Light Reflection, Cameraorientation, And Uniformity Of Eye Shape Were Identified. Testing Data Underwent Pre-Processing Andfeature Extraction Following The Same Methodology As Model Training.Experimental Research Approaches Were Adopted To Address The Research Questions, With The proposed Iris Segmentation System Consisting Of Localization, Normalization, Feature Extraction, Andmatching Components. The Importance Of Relevant Feature Extraction In Building An Efficient Irisrecognition Model Is Highlighted, Utilizing Techniques Such As HOG For Structure Description Andvector Support Machines For Classification. The Unet Model Has Been Used To Segment Iris Datasets Obtained From The Cassia Repository, Utilizingspecific Parameters. Results Indicated That Efficient Iris Segmentation Presents Challenges Dependenton Dataset Specificity And Feature Extraction Techniques. Thus, Researchers And Stakeholders Areencouraged To Prioritize Feature Extraction And ROI Identification To Develop Efficient Irisrecognition Models.Challenges In Iris Segmentation, Particularly Related To Complex Eye Structures, Were Observed, Withpotential Solutions Including The Use Of Standard Cameras For Clear Image Acquisition And Mitigatingthe Impact Of Light On Pupil Visibility. In Conclusion, Further Exploration In Biometric Securitytechniques Is Warranted To Design Robust And Generalized Iris Recognition Systems. Youngresearchers Are Encouraged To Innovate In This Domain To Safeguard Organizational And Personalresources Effectively Using Iris Recognition Methods.