roi_sort

roi_sort processes Region of Interest (ROI) files to extract features from finger vein images and enable biometric identity verification.

Key Features:

• ROI file processing: Processes and manages Region of Interest (ROI) files for downstream biometric analysis.
• Feature extraction with (2D)²PCA: Applies (2D)²PCA (Two-Dimensional Principal Component Analysis) to extract discriminative features from finger vein patterns.
• Metric learning and KNN classifier: Uses metric learning to train individualized K-Nearest Neighbors (KNN) classifiers for personalized identity verification.
• Class-imbalance handling with SMOTE: Integrates Synthetic Minority Over-sampling Technique (SMOTE) to generate synthetic samples and mitigate class imbalance.

Scientific Applications:

• Finger vein recognition: Verifies identities using unique finger vein patterns for biometric authentication, with reported recognition rates up to 99.17%.

Methodology:

(2D)²PCA for feature extraction, metric learning to create individualized KNN classifiers, and SMOTE for synthetic minority over-sampling applied to ROI files.