Research On The Key Technology Of Finger Vein Recognition For Fuzzy Image

As people pay more and more attention to the security of personal identity information,personal identity identification plays an important role in our daily life.People seek a more convenient,safe and fast identification method--biometric identification.At present,there are more and more scenes of biometric recognition technology in the market,such as face recognition and check-out in unmanned restaurants,fingerprint gate at the community gate,iris unlocking on mobile phones,finger vein deposit and withdrawal in ATM machines,etc.Among them,finger vein recognition technology,which uses human internal finger vein as a biological feature,has the advantages of fast identification speed,high accuracy,in vivo detection,characteristics are not easy to be copied.The problem of low security and poor usability of traditional biometric identification technology is solved effectively.In the existing finger vein recognition system,the problem of poor recognition performance of fuzzy image is hard to avoid.There are two main reasons for fuzzy images: external factors,such as illumination,temperature,exposure uniformity,etc.Lead to poor image quality and unclear venous structure during imaging.Placement factors: such as users' habits,placement attitude,etc.Resulting in large image deviation or virtual focus and other problems.There are some problems such as poor contrast and less effective vein information in fuzzy vein images,which leads to inaccurate feature extraction and poor matching and recognition performance.Therefore,solving the problem of fuzzy images has become the focus of domestic and foreign researchers.The research work of this paper is mainly divided into the following parts:(1)This paper introduces the existing two kinds of finger vein recognition techniques : refers to the vein recognition based on feature points matching method and gray texture matching method based on image registration,which introduces the MHD recognition method based on the fine line distance,LBP partitioned histogram recognition methods and multiple directions LLBP recognition principle of identification method and main process.According to the characteristics of fuzzy image,the shortcomings of the three algorithms are analyzed respectively.(2)A dynamic digital vein recognition method based on image pyramid model was proposed.First of all,after classifying the images by fuzziness by using the fuzzy detection function of the image pyramid model,a method using dynamic parameter Ni Black segmentation and dynamic threshold recognition is proposed.Compared with MHD recognition method based on thin line distance,this method first carries out dynamic parameter Ni Black segmentation according to the image blur degree,effectively solving the problem of more pseudo-veins and inaccurate featureextraction caused by single segmentation threshold when extracting the features of fuzzy images.Furthermore,by setting the dynamic recognition threshold,the identification performance loss of partial clarity comparison can be avoided,so that the identification performance of the whole system can be improved greatly.(3)A weighted nearest neighbor binary model method based on top-k block segmentation is proposed,which improves the image recognition process with a large degree of blur.Compared with LBP partitioned histogram recognition methods and multiple direction LLBP recognition method,this method by extracting image grayscale texture characteristics of horizontal and vertical direction,using the finger vein structure to give priority to with vertical,horizontal extension of the distribution characteristics of less,is obtained by weighted statistics,effective method to solve the traditional LBP texture feature extraction is not accurate.Furthermore,the selection of top-k blocks can effectively eliminate unnecessary interference parts,and solve the error problem of noise introduced by LBP method and multi-directions LLBP method in the feature comparison of the full graph,so as to further improve the accuracy of matching and identification.(4)Finally,this article proposed based on dynamic image pyramid model refers to the vein recognition method and weighted based on top-k block nearby for a binary pattern method fusion recognition system,by comparing the MHD recognition based on thin line distance method and the single dynamic identification method,after experimental verification,to prove after fusion recognition system for the existence of blurred image refers to the vein recognition system has a better recognition performance,and more practical.