Research On Feature Extraction And Fast Recognition Of Finger Vein Image

Finger vein recognition is a new biometric recognition technology that has significantly improved uniqueness,stability,portability,and bioassay characteristics,meaning that it has wide application and development prospects.In this paper,finger vein near infrared image pretreatment,vein feature extraction,matching recognition algorithm,rapid retrieval strategy are studied.The main contents are follows:(1)A new method of modular adaptive histogram equalization(MAHE)is presented to enhance image contrast and reduce computational cost.The experimental results illustrated that the recognition accuracy of MAHE and AHE were close,while MAHE consistently outperformed AHE in calculation time,through respectively comparing the receiver operating characteristics(ROC)curves and cumulative match characteristics(CMC)curves.(2)An adaptive image binarization method based on effective information loss ratio and interference lines elimination rate is proposed.It can adaptively obtain segmentation threshold to adapt the vein image of different illumination intensities and image quality.For the problems during feature extraction,such as losing useful information or obtaining incomplete information,image fusion method based on morphological algorithm and maximum curvature extraction is studied.(3)The Hausdorff average distance matching algorithm and templates overlap rate matching algorithm are studied.We used the Hong Kong Ploytechnic University finger vein image database to carry out the experiment.The matching performances are analyzed through FAR,FRR,ROC,EER.The advantage,disadvantage and application condition of two matching algorithm are researched.(4)There are some problems caused by large amount of image library during matching identification,such as large storage space,large matching computational cost,poor recognition effect.A joint algorithm combined image compression with image encryption is proposed to reduce the size of characteristic template and protect users' privacy.Based on secondary characteristic of image,a rapid retrieval strategy is presented to improve the recognition efficiency.