DESIGN OF A CNN-BASED IRIS BIOMETRIC FRAMEWORK FOR UNDER-AGEVOTER SCREENING: INTEGRATION OF EXPERIMENTAL DESIGN, MACHINE LEARNING PIPELINES, AND OBJECT-ORIENTED MODELLING

Abstract

This article synthesizes the architecture and design contribution of a doctoral study proposing an iris-image convolutional neural network (CNN) to screen for possible underage voter registration. The work integrates experimental design, a reproducible machine-learning pipeline, and object-oriented software modelling. The proposed input–preprocessing–feature-learning–classification sequence includes quality screening, normalization, convolution/ReLU blocks, max pooling, flattening, dense feature integration, and two-class softmax inference. A human-review pathway is incorporated to prevent an uncertain biometric prediction from being treated as proof of legal age. The source thesis provides detailed architectural descriptions and worked performance calculations, but the available evidence does not independently verify population-level age discrimination from iris texture. The principal contribution is therefore a specified, testable research architecture and a validation protocol, not demonstrated operational eligibility determination. Recent literature on iris recognition, segmentation, generalization, and responsible AI is used to identify required controls before deployment.