Improved Fuzzy Clustering (IFC) and Correlation Based user Threshold Selection with TRI-Branch for Finger Vein Recognition

Abstract

During the process of template matching with regard to finger vein identification, the probe used will get permitted only when the count of its vein points overlapping with the registered client is higher compared to the predetermined threshold. But, the admittance might be incorrect due to the neglect of the structure of the vein pattern. In the earlier works, the extraction of the vein structure (tri-branch) is done out of the vein pattern, and then combined with the entire vein pattern using a user-oriented threshold dependent filter setup. It renders a greater value of false acceptances, due to the User-oriented Threshold obtained from filter. In the step of branch tracking, the closest points are falsely detected, and therefore few tracking algorithms are needed. In order to resolve this, user-specific Threshold is generated on the basis of the correlation filter based selection combined with genetic algorithm. In the step of branch tracking, the closest points between the samples are decided by using the improved fuzzy clustering algorithm. It is observed that the local branches of the vein close to the segregation point of the vein pattern differ hugely from the fake pictures. The results of experiment carried out on two publicly available databases show the efficiency obtained of the novel design for boosting the performance achieved with respect to vein pattern based finger vein identification.