Robust Fingerprint Identification Using Canny and Sobel Edge Enhancement

Abstract

Fingerprint recognition has long been considered one of the most reliable biometric methods due to the uniqueness and consistency of fingerprint patterns. However, real-world scenarios often present challenges such as noise, smudges, partial prints, or deliberate adjustments that can hinder accurate identification. This study presents a hybrid approach to fingerprint recognition that combines fine detail features and large-scale static feature conversion descriptors to improve flexibility and accuracy. To enhance the clarity of fingerprint patterns, preprocessing techniques such as Canny edge detection and Sobel filtering are applied, enabling better feature extraction. Detailed points are taken out using the transit number way and authorized by filtering the area, reducing the mass, and analyzing the structure of the ridges. Sort descriptors are used to capture local texture information and are normalized to ensure uniform vector lengths. An audit dataset, comprised of original and synthetic fingerprints with different degrees of complexity, is used to validate the system. The combination of fine details, proofing features, with careful pre-processing obtained the best classification rate of 96% across tested methods. The proposed approach achieves good results on the identification of variable fingerprints with the performance efficiency required for real world, real time, and resource constrained forensics.