Design And Implementation Of Iris Capture Device And Research On Several Problems In Iris Recognition System

With the development of information technology, security requirements ofinformation technology is growing.In this situation,the traditional password orauthentication technology has been difficult to meet people’s security needs.At the sametime, biometric technology,as a kind of authentication technology,with high reliability andease of use,well meet this demand.And in recent years,biometric technology has beendeveloping rapidly.Among them, iris recognition technology has gradually become a hotfield,because of its outstanding advantages.A complete iris recognition system includes thefollowing procedures:Iris capture, iris image quality assessment, iris preprocessing, irisfeature extraction, iris feature matching.a) Iris capture：The basic theory of the iris caputre was introduced firstly.Then we analysedthe hardware-related factors affecting the quality of the caputred iris image.We foundthat the focal length and material of the lens,the type of the light and the quantumefficiency of the sensors highly correlated with the quality of the captured iris image.Based on the analysis of several existing iris acquisition mode,we found that thenon-contact&close mode is the best mode which takes not only cost but also the ease ofuse and the difficulty into consideration.Accordingly,this paper proposed a completesolution of iris capture device which has the followingcharacteristics:non-contact,excluding the influence of rotation,excluding the naturallight,high ease of use,low cost.b) Two traditional iris location algorithms were introduced fistly.Based on the analysis ofthe advantages and disadvantages on them,we found that these two algorithms havedifferent advantage in computation speed and location accuracy.Accordingly，a newlocation algorithm,combining the advantages of these two traditional algorithms,wasproposed in this paper.And the experiments,based on library CASIA V1.0and libraryCASIA V3.0_INTERVAL,demonstrateed that the proposed iris location algorithmimproved in both speed and accuracy.c) Firstly, the normalized segmented iris was divided into multi tracks and each track wasestimated individually by Gabor filter and SVM to analyze RAR.We found thatdifferent tracks has different result of RAR.Accordingly,two local quality evaluationparameters were used to analyze each track of the iris image,getting weightedcoefficients of different tracks. Finally, all tracks’ information was fused according tothe weighted coefficients of different tracks. The experiments,based on library CASIAV1.0and library CASIA V3.0_INTERVAL,demonstrateed that the proposed algorithm,based on weighted fusion strategy,can effectively improve the performance ofiris recognition system.To sum up,this paper has finished the following works:1) We proposed a complete solution of iris capture device.2) We proposed an iris location algorithm which combines two traditional algorithms.3) We proposed an iris matching algorithm which is based on weighted fusion strategy.