Development of an Enhanced Giant Armadillo Optimizer Technique for Fingerprint Recognition System

Abstract

Fingerprint recognition systems continue to experience significant challenges due to variations in image quality, skin deformation, partial impressions, environmental noise, and the increasing dimensionality of fingerprint databases, all of which negatively affect recognition accuracy and computational efficiency. Metaheuristic optimization techniques have therefore gained attention for feature selection, with Giant Armadillo Optimisation (GAO) demonstrating strong capability in balancing exploration and exploitation during the search process. Nevertheless, GAO still suffers from issues such as slow convergence, parameter sensitivity, and local optima stagnation, thereby creating the need for enhanced GAO models. Hence, this study developed an enhanced giant armadillo optimized feature selection technique for fingerprint recognition system. Fingerprint samples were collected from 500 students of LAUTECH, Ogbomoso, using a Digital Persona fingerprint sensor, with three samples obtained from each participant to ensure data reliability and variability. The second stage, image pre-processing, involves a sequence of enhancement operations such as image enhancement, noise removal, segmentation, binarization, and thinning to improve image clarity and ridge structure. The third stage, feature extraction, focused on identifying key fingerprint characteristics, where minutiae extraction performed using the Crossing Number (CN) algorithm to detect ridge endings and bifurcations. In the fourth stage, feature selection carried out using the formulated Enhanced Giant Armadillo Optimizer (EGAO) integrated with Adaptive Parameter Tuning (APT) to optimally select the most discriminative and relevant features for recognition. The fifth stage involved system implementation in MATLAB R(2020a), where all processes from pre-processing to feature selection and classification programmed and executed. Finally, the sixth stage focused on performance evaluation of the developed fingerprint recognition system using metrics such as False Positive Rate (FPR), Sensitivity, Specificity, Precision, Accuracy, and Recognition Time. The results of the EGAO achieved a False Positive Rate (FPR) of 3.70%, sensitivity of 97.84%, specificity of 96.30%, precision of 98.75%, and accuracy of 97.45%, and recognition time of 78.86 seconds. In comparison, the GAO had an FPR of 12.96%, sensitivity of 94.75%, specificity of 87.04%, precision of 95.64%, and accuracy of 92.82%, with recognition time of 41.91 seconds. The study demonstrated that EGAO consistently outperformed GAO in false positive rate, sensitivity, specificity, precision, and accuracy, confirming that APT enhanced the optimization capability of GAO and enabled the selection of more discriminative fingerprint features. Although EGAO incurred a higher recognition time than GAO, its substantial improvement in recognition performance establishes the EGAO framework as a more accurate, reliable, and effective approach for fingerprint recognition and robust biometric identification applications.