Classification of Finger Vein Image Using Convolutional Neural Network

Currently available technologies can perform rather well, but their effectiveness is mostly contingent on how well the venous images being analyzed are quality images. Finger vein features have garnered significant attention in the past few years as a potential means of automatic user identification. A significant amount of daily usage goes into the very vital personal identification procedure. The identification process is applicable in the workplace, private zones, and banks. Humans could be a rich topic having abundant features that may be used for identification purposes such as finger veins, iris, and face. This research proposes a Convolution Neural Network (CNN) based two-stage finger vein classification and identification method and discusses the model performance with four methods of extracting features such as Gabor, Speeded Up Robust Features (SURF), Local Binary Patterns (LBP) and Principle Component Analysis (PCA) and comparing the results of the proposed classification system with another classification method Feed Forward Neural Network (FFNN). The experiment is conducted on images acquired from a lot of subjects of the Sains Malaysia database to illustrate the performance of the proposed algorithm. The result shows a superior performance to the convolution neural network of biometrics in the proposed system and shows the LBP features extraction method outperforms the other methods such as (Surf, Gabor, and PCA).