Recognition Algorithm Of Finger Vein Image

With the rapid development of information technology, people pay more and more attention toinformation security. Personal identification is an important method of information securityprotection. Biometric recognition uses the innate and unique human body characteristics for thepersonal identification, which provides more safety and more convenience than the traditionalidentity recognition method. In the current biometric recognition methods, finger vein recognition isa method utilizing the vein patterns inside one’s fingers for personal identification, which has manymerits, such as living body recognition, interior characteristic, non-contacting and so on. Therefore,the finger vein recognition has become a very important research field in biometric recognition.However, in practical applications, the captured finger vein images usually have low qualitybecause of the influence of the used equipment, and it is difficult to extract the features from thoselow quality finger vein images. To address the problem, we do some research on the finger veinimage recognition. In this paper, we propose a novel finger vein image recognition method based onevolving finger vein template using OPM and based on subspace using two-directional combinationof two-dimensional kernel principal component analysis. The main attributions in this paper are asfollows:1. This paper introduces the gathering way of finger vein image using the reflected light andtransmitted light, and introduces its pre-processing, including the extraction of its ROI, scale andgray scale normalization, image segmentation and finger vein image feature extraction, imagepost-processing. Finally, we introduce two kinds way of matching recognition, and the performanceparameters to judge the system performance.2. This paper proposes a new method of templates evolution. We can generate a fused templateas registered template of each finger form a number of templates after its being pretreated, whichcan be called representative template. Then, we match it with the testing templates for identification,and obtain various identification parameters. Studying shows that the number of templates forfusion is not the more the better, so we introduce how to select the best number of templates. At thesame time, we find that it does not conform to the actual identification system treating each pixel oftemplate equally. So we put forward the OPM to weight the value of registered template in differentlocations. The experimental result shows that finger vein image recognition method based onevolving finger vein template using OPM can evalulate the performance of system recognition.3. In the part of feature matching method based on subspace, we introduce several methods oftwo-dimensional linear analysis, and have realized two-dimensional principal component analysis(2DPCA), two-dimensional maximum margin criterion (2DMMC), two-dimensional Fisher linear discriminant analysis (2DFLD) and combination of these two-dimensional linear analyses. Inaddition to, the main content and steps of two-dimensional kernel principal component analysis(K2DPCA) are given in detail. Meanwhile, we propose a new recognition algorithm based ontwo-directional two-dimensional combination of kernel principal component analysis. Through theexperiment, we have realized K2DPCA+2DPCA、K2DPCA+2DMMC、K2DPCA+2DFLD andget their recognition rate. The experimental results show that two-directional two-dimensionalcombination of kernel principal component analysis has a better recognition effect.