A Biometric-Finger Vein Authentication System for Security Purpose using Deep Learning Technique

This research presents a Biometric-Finger Vein Authentication System designed for enhanced security applications, employing a Deep Learning Technique. The proposed system undergoes a series of image processing steps, starting with the conversion of the input finger image to grayscale. Subsequent stages involve applying a Gaussian filter, amplifying the vein structure, performing Histogram Equalization, Canny edge detection, skeletonization, segmentation, perimeter extraction, and defining the Region of Interest (ROI). Feature extraction is conducted, and the extracted features are matched using Convolutional Neural Network (CNN) classification. The classified features are then compared to the finger vein data stored for registered users in the database. The authentication process results in a determination of either "authentication successful" if a match is found, or "authentication failed" in the absence of relevant data. The system's accuracy is evaluated based on the success or failure of this matching process, emphasizing its efficacy in reliably authenticating individuals through the analysis of finger vein patterns.

Keywords: Finger Vein Authentication, Deep Learning, CNN, Preprocessing, Segmentation, ROI Location, Feature Matching, Classification and accuracy.