Research Of Finger Vein Recognition Technology Based On Matrix Factorization

In recent years,with the identity authentication technology becoming more and more popular in people's daily life,biometric technique is widely used in security,education,payment and other fields.As a new biometric technology,finger vein recognition has attracted much attention.It has the advantages that common biometrics do not have.For example,the characteristics of finger vein itself are living,high security,not easy to be forged and so on.Finger vein recognition consists of three parts:region of interest extraction,feature extraction and recognition.The quality of finger vein images are easily affected by light and position,which is not conducive to feature extraction.Therefore,to explore a reasonable image preprocessing method is the basis of the recognition.Feature extraction is the most important part of recognition.In the field of image recognition,the images are stored and represented in the form of matrix.Matrix factorization is an important method to extract local or global features,and it is also an effective means to achieve the dimension compression.How to propose an effective feature extraction and identity recognition scheme based on the original matrix factorization algorithm is the focus of this paper.The main research works of this paper are given as follows:(1)A method to extract the region of interest from the digital vein based on the rotation correction is proposed.During the collectionprocess,the original image will contain random noise and deflection,so it is necessary to accurately extract the most concentrated area of the feeature of the finger vein.In this method,the rotation angle is determined by finding the straight line of the finger edge,and the region of interest is accurately extracted by drawing the histogram of pixel gray levels in row direction and column direction.The feasibility of the proposed scheme is verified by theoretical and experimental experiments.(2)The feature extraction algorithm of local binary pattern fusion principal component analysis and the two-dimensional non-negative matrix decomposition based semi-tensor product are proposed.The traditional matrix decomposition algorithm will lose part of the local structure information or two-dimensional structure information of the image.So before feature extraction,the local binary mode processing is used to process the images.Experiments show that the fusion of local binary patterns can effectively improve the recognition rate of traditional matrix factorization,and the semi-tensor product can reduce the factorization complexity and improve the factorization efficiency.(3)A system architecture of finger vein recognition is proposed.Using the image preprocessing and feature extraction method proposed in this paper,the finger veins are recognized by support vector machine to achieve the purpose of user identity authentication in 1:N environment.