An Efficient Hand Dorsal Vein Recognition Based on Neural Networks

Abstract

This paper proposes an effective human hand vascular pattern recognition system by using multi-layered perceptron neural networks for biometric identification applications. Biometric hand dorsal vein images are acquired by using NIR (near infra-red) illuminated CCD camera from 103 persons of different ages and gender. The vascular region of interest (ROI) is cropped from hand images firstly and then mean filter and histogram equalization processes were applied to the 240x180 pixels resolution hand vein pattern images in order to restrain the noises. The gray-scaled vein pattern images were converted to the binary format by applying Otsu’s thresholding method. The resulting images then are divided into 20x20 pixel dimensioned sub-images before feature extraction. Average absolute deviation (AAD) algorithm was implemented to these sub-images for getting the feature sets. Multi-layer perceptron neural network (MLPNN) method was performed for identification of the human hand vein pattern images. Experimental results showed that the proposed method achieved correct classification rates up to 100%