A New Personal Verification Technique Using Finger-Knuckle Imaging

Hits: 82
Dyscyplina naukowa:
Rok:
2016
Typ publikacji:
Artykuł
Słowa kluczowe:
Biometrics, Finger-Knuckle imaging, Cross-correlation, k-NN
Autorzy:
Czasopismo:
Computational Collective Intelligence
Strony:
515–524
ISSN:
978--3-31
Punkty MNiSW:
15
Streszczenie:
This paper focuses on automatic pattern-based extracting of biometric features where finger-knuckle images are analyzed. Knuckle images are captured by digital camera, and then by the image processing techniques the most relevant features (patterns) are discovered and extracted. Knuckle-based images were filtered by the Hessian filters. It enabled to enhance image regions with image ridges. In the next stage similarity of images were computed by the Normalized Cross-Correlation algorithm. Ultimately, similarities were classified by the k-NN classifier. The discovered features belong to so-called human physical features, which involves innate human characteristics. Physical biometric features can often be gathered with specialized hardware, needing only software for analysis. That capacity makes such biometrics simpler. We conducted a variety of experiments and showed advantages and disadvantages of the approaches with promising results.

Copyright © 1997-2012 - Zakład Komputerowych Systemów Biomedycznych.
Uniwersytet Śląski, Wydział Informatyki i Nauki o Materiałach, Instytut Informatyki, ul. Będzińska 39, 41-200 Sosnowiec, Polska