Fingerprint Authentication System with Feature Detection using Descriptor Analysis

Fingerprint identification is a fundamental biometric modality widely utilized in various applications such as law enforcement, border control, access control, and financial transactions. The uniqueness and permanence of fingerprints make them an ideal candidate for reliable and secure personal identification. The processing speed and special hardware configuration are the main computational constraints for any Fingerprint authentication system as it limits its computational capability. The SIFT and SURF both are suffered from these limitations due to size of descriptors and key points localizations. In this research study the descriptor-based feature matching for fingerprint identification is studied. the SIFT and Orb techniques for feature extraction, descriptor generation for extracted features from fingerprint taken from FVC2002 DB1b database, post processed it using PCM variant for dimensionality reduction is exploited , and then performed similarity measurement using Euclidean distance, with the aim of improving the accuracy, and scalability of recognition. Through comprehensive experimentation and analysis, it is aimed to demonstrate the efficacy of descriptor-based approaches in enhancing the performance of fingerprint matching. The Precision-Recall and ROC curves have demonstrated that the PCA based descriptors are far more distinctive in identifying the fingerprint as compare to ORB and SIFT based methods.