Zařízení pro snímání a rozpoznávání cévní struktury prstu

Abstract

This thesis is focused on the design and implementation of a device for capturing the vascular structure of the finger using near-infrared (NIR) imaging and the subsequent processing of image data for biometric identification purposes. A specialized imaging device was developed to acquire multi-view images of the finger, which were subsequently preprocessed using methods such as Gabor filtering, contrast limited adaptive histogram equalization (CLAHE), and morphological operations. Following preprocessing, the vascular structure was segmented using Otsu's thresholding method. The similarity between the segmented images was analysed using a correlation-based method, demonstrating that images with a small rotational difference exhibit a high degree of similarity, while images with greater rotation show a decrease in similarity. The results confirm the potential of multi-view imaging for enhancing biometric identification and simultaneously highlight the limitations of 3D reconstruction based solely on segmented data.