<![CDATA[Digital access security requires authentication mechanisms that are more reliable than conventional passwords, PINs, identity cards, and user IDs, which are prone to being forgotten, guessed, copied, or misused. This study develops a biometric authentication system based on dorsal hand vein patterns by combining texture feature extraction and the Multiclass Support Vector Machine (SVM) method. Dorsal hand veins were selected because they are unique to each individual, located beneath the skin surface, difficult to forge, and suitable for contactless acquisition. The dataset consisted of 220 dorsal hand vein images from five male participants aged 18–25 years, captured using a Near-Infrared (NIR) LED camera. The data were divided into 80% training data and 20% testing data. Image preprocessing included resizing, cropping, grayscale conversion, Top-Bottom Hat Transformation, CLAHE, Median Filter, morphological smoothing, ROI segmentation, erosion, and dilation. Texture features were extracted using a Gabor Filter with a wavelength of 3 and an orientation of 135°, producing mean, variance, and entropy values. Classification compared Linear, Polynomial, and RBF kernels, with parameter optimization using GridSearchCV. Evaluation using 10-Fold Cross-Validation showed that the RBF kernel with C = 1 and γ = 4 achieved the best performance, with 88.89% accuracy, 89.07% sensitivity, and 97.23% specificity, indicating the feasibility of this approach for dorsal hand vein biometric authentication. Further testing with larger datasets is still required.]]>
Copyrights © 2026
