An algorithm for determining SIFT matching score for dorsal vein recognition system

Abstract

Vein pattern recognition has been an increasingly biometric branch nowadays. This technology has many advantages over other biometric technologies. This technology works only on alive person. Vein pattern of every person is unique even in case of twins. Whole procedure of registration for verification of a person gives a hygienic opportunity. Vein pattern of a person remains constant throughout the life until and unless, physically damage. Recognition through veins cannot be affected from aging, color and physical environment because veins are present underneath the skin. NIR cameras of wavelength of 700 nm to 1000 nm are used to capture the images of vein patterns. When infrared radiation falls on the veins, these get illuminated in dark color due to the absorption of radiation by the hemoglobin present in the veins. The SIFT (Scale-Invariant Feature Transform) algorithm has given very good results in feature extortion and matching but it does not provide matching score of features. In this work, we proposed an algorithm to determining the matching score of pattern matched through SIFT algorithm. For experimental purpose, we performed image acquisition, pre-processing, feature extraction and filtering to eliminate noise from the images. We tested our algorithm on the database of 160 persons and we calculated performance of our algorithm in terms of FAR and FRR 2.7% and 4.5% respectively.