Hand Vein Region of Interest (ROI) Extraction Using Faster Region-Based Convolutional Neural Network (R-CNN)

The extraction of the region of interest (ROI) in hand vein images plays a crucial role in their detection. Accurately extracting the vein area faces challenges such as variations in hand poses, lighting conditions, orientation, appearance, and noisy background. Although numerous techniques have been proposed for ROI extraction of hand veins, their capabilities are often limited to a specific hand pose and location. To address this limitation, this paper presents a deep learning- approach using Faster region-based convolutional neural network (R-CNN) for adaptive ROI vein extraction. The proposed system is evaluated using two hand vein databases: self-acquisition and Sakarya University of Applied Sciences (SUAS), encompassing diverse hand poses. To assess the performance of the proposed technique, it is compared to two existing ROI vein extraction techniques. The comparative results demonstrate that the proposed technique achieves impressive performance in accurately locating the ROIs for various hand poses and locations. By employing a deep learning-based approach and evaluating its effectiveness on different hand vein databases, this appear offers a promising solution for adaptive ROI vein extraction, overcoming the limitations of existing techniques.