SIFT-based Approach for Dorsal Hand Veins Images Recognition

In this paper we present a system for person identification using dorsal hand veins pattern images. A simple approach is developed based on scale-invariant feature transform (SIFT) method for image features extraction, fast approximate nearest neighbors algorithm (FLANN) for keypoints matching in a preselected region-of-interest (ROI), and subsequent utilization of dense layer neural network for accurate friend-or-foe predictions. The efficiency of the considered approach is evaluated using an open dataset, and confusion matrices are calculated. The obtained results confirm again that dorsal hand veins identification systems are particularly promising for high-security settings. The proposed approach is implemented in software for Orange pi single-board computers.