No abstract

Biometric Recognition Systems allow individuals to be automatically authenticated or identified by using their unique characteristics. Finger vein (FV), widely used for this purpose, has a crucial place among biometric systems because of its advantages, which are user-friendliness, ability to detect living tissue, high reliability, low system cost, and less area requirement in installation. It has a wide usage area, especially in places where personal safety is at the forefront. In this study, we examine the effect of the Horizontal and Vertical Total Proportion (HVTP) feature extraction algorithm on the success rate when the fusion technique is applied. Homomorphic Filter (HF) and Perona-Malik Anisotropic Diffusion (PMAD) are used to remove the noise and light scattering issue in the FV databases, and Gray Level Run Length Matrices (GLRLM), Gray Level Co-occurrence Matrices (GLCM), Segmentation-based Fractal Texture Analysis (SFTA), Horizontal Total Proportion (HTP), and Vertical Total Proportion (VTP) methods are applied to describe texture features. The fusion of multiple features instead of using only one type of feature can improve the accuracy of FV recognition systems. The novelty of the study is the fusion of HTP and VTP with the GLRLM, GLCM, and SFTA features by using Yang finger vein databases (Database_1) and MMCBNU_6000 (Database_2). Experimental results reveal that the HTP and VTP significantly improved the classification success in these FV image databases. The best success rate achieved in the Ensemble classifier is 99.7% using Database_1 and 97.6% using Database_2.