Research On The Key Problems Of Finger Vein Recognition

With the rapid development of information technology, information security is paid more and more attention to by the public widely. The disadvantage of primitive identity authentication methods is emerging gradually. Biometric identification technologies which are safer and more reliable arise under this environment. Among them finger vein recognition technology has been a research focus relying on its advantage including living body recognition, internal features, non-contact and high security level. Although companies and scholars have achieved gratifying results in recent years, there are key problems which badly need to be figured out.This thesis does research on the key problems of finger vein recognition. The main work and research achievements can be summarized into four aspects as follows.Firstly, this thesis designs the finger vein image preprocessing method which is suitable for the finger vein image database adopted by this thesis. A finger region localization method based on gradient and gray information is proposed. This thesis also proposes ROI (region of interest) extraction method based on finding the median location of gray mean of right side finger section. The preprocessing method is turned out to be effective by further processing;Secondly, A finger vein restoration algorithm based on removing the skin scattering is proposed. This algorithm analyzes the reason of finger vein image blurring. It estimates the PSF (point spread function) of blurry finger vein image adaptively using the prior knowledge of skin optical. It achieves the clarity of finger vein image excellently;Thirdly, A growing type finger vein segmentation algorithm based on SVM is proposed. This algorithm selects the seed points using multi directions vein section features. It builds the SVM growing criteria model using supervised learning. It can resist the influence of noise and shadow during segmenting finger vein.Finally, A finger vein recognition method based on vein bifurcation point calibration and SIFT features is proposed. The experiment proves that the target should be the finger vein image that has been segmented when extracting SIFT features. The experiment data proves that calibrating finger vein image before SIFT feature matching can significantly reduce the EER (equal error rate) of finger vein recognition system at the cost of a small increase in recognition time.