Robust contactless fingerprint authentication using dolphin optimization and SVM hybridization

The field of contactless fingerprint (CLFP) recognition is rapidly evolving, driven by its potential to offer enhanced hygiene and user convenience over traditional touch-based systems without compromising security. This study introduces a contactless fingerprint recognition system using the Dolphin Optimization Algorithm (DOA), a nature-inspired technique suited for complex optimization tasks. The Histogram of Oriented Gradients (HOG) method is applied to reduce image features, with DOA optimizing the feature selection process. To boost prediction accuracy, we fused the DOA with a Support Vector Machine (SVM) classifier, creating a hybrid (DOA-SVM) that leverages the global search prowess of DOA alongside the reliable classification strength of SVM. Additionally, two more hybrid models are proposed: one combining Fuzzy C-Means (FCM) with DOA-SVM, and another combining Neutrosophic C-Means (NCM) with DOA-SVM. Experimental validation on 504 contactless fingerprint images from the Hong Kong Polytechnic University dataset demonstrates a clear performance progression: DOA (91.00%), DOA-SVM (94.07%), FCM-DOA-SVM (96.03%), and NCM-DOA-SVM (98.00%). The NCM-DOA-SVM approach achieves superior accuracy through effective uncertainty handling via neutrosophic logic while maintaining competitive processing efficiency. Comparative analysis with other bio-inspired methods shows our approach achieves higher accuracy with reduced computational requirements. These results highlight the effectiveness of combining bio-inspired optimization with traditional classifiers and advanced clustering for biometric recognition.