Методи сегментації вен долоні у системах біометричної ідентифікації

Abstract

This work is aimed at researching and improving palm vein image segmentation for secure, non-contact biometric identification systems. The primary objective is to isolate subcutaneous vascular networks from complex backgrounds, overcoming inherent challenges such as low-contrast imaging, asymmetric illumination, and skin surface contamination. The study evaluates traditional image enhancement techniques, maximum curvature methods, alongside advanced deep learning architectures that integrate multi-scale convolutions and Swin-Transformer networks