Özet:

Abstract Palmprints are stable and unique information used for biometric identification. This paper devises a novel texture-based feature extraction method inspired by the Gammadion structure that recognizes an individual based on Palmprint. The Region of Interest (ROI) extracted from the palm is initially normalized using the Histogram equalization technique. The ROI is converted into the frequency domain using Shearlet transformation to represent it in an illumination invariant form. Then the gammadion structure-based feature extraction method is used. It generates multiple feature maps fed to a simple Convolutional Neural Network(CNN). The proposed methodology is invariant with illumination and noise. Four publicly accessible standard Palmprint databases (CASIA, IITD, Tongji, and PolyU2) are used for extensive experimentation. The result is then compared with existing state-of-art techniques. The experimental analysis shows that the proposed methodology obtains the highest accuracy of 99.45% for the PolyU2 dataset, which is superior to some existing methods in the literature.

Anahtar Kelimeler: