Vein Uncovering And Recognition Of Visible Light Image

At present,most identification verification systems rely on human's fingerprints,palmprints and face recognition technique based on visible light or structure light.Nevertheless,biometrics mentioned above do have apparent defects which narrowed their application scenario sometimes.For example,we use our hands so frequently that they are easy to get injured or stained,this will affect the discriminative features such as fingerprints and palmprints negatively.For existing visible light face recognition technologies,researchers have approved that the visible light face characteristics are vulnerable as they might be changed due to the aging or face-painting.And for structure light-based face recognition methods,it is particular difficult to differentiate similar individuals in some cases,for example,a pair of identical twin.Recently,studies on vein biometric are flourishing.On the one hand,the vein pattern is intrinsic and enduring.Also,the vein pattern is hypodermal and protected by the skin.However,many of existed vein pattern recognition methods need near-infrared or laser sensors to collect the image,which requires specific complex equipment.To enlarge the application scenario of vein recognition,we proposed a GAN based uncovering method to extract vein patterns of visible light image effectively.Sometimes,images may suffer from motion blur,to enhance the performance under motion-blur environments,we developed a fully-convolutional network based deblurring algorithms to recover relatively sharp image before extracting the vein patterns.Finally,we designed a powerful feature extractor to implemented identification recognition system using vein pattern,the system showed great performance in our experimental estimations.