Publication: İris görüntülerinin analiziyle kimlik tanıma
Abstract
İRİS GÖRÜNTÜLERİNİN ANALİZİYLE KİMLİK TANIMA Günümüz teknolojisinde ek güvenlik ihtiyaçlarına cevap verebilen biyometrik sistemler, kişilerin ve biyolojik ürünlerin sahip olduğu, özgün fiziksel ve davranışsal özellikleri kullanan bir örüntü eşleme sistemidir. Biyometrik sistemler, şifre hatırlama ya da kimlik taşıma gibi ihtiyaçlara gerek duymadan, sadece kişiye ait olan özellikleri kullanır. Bu sistemler kişiye ait bilgileri kaydedip bir veritabanı oluştururlar ve daha sonra her güvenlik sorgulamasında kullanırlar. Bu tez kapsamında yaşam boyu kararlı olması, ayrıt edici özelliklerinin diğer biyometrik tekniklere göre çok daha fazla olması, iyi korunan iç organ olması,taklit edilmesinin imkansız olması,uzaktanda görülebilmesi, yüksek rastgeleliğe sahip olması, genetik benzerliğin azlığı, görüntü alınırken herhangi bir temas olmadan kolayca alınabilmesigibi nedenlerden dolayı irisi temel alan bir biyometrik tanıma sistemi üzerinde çalışılmıştır. Sistemde kullanılan veritabanındaki imgelerden iris bölgesinin elde edilebilmesi için, Canny kenar bulma algoritması ve Hough transformu kullanılmıştır. Alınan göz imgelerinde kestirilen iris imgesi kutupsal formdan dikdörtgen forma dönüştürülmüştür. Normalleştirilmiş iris deseni Gabor dalgacıkları ile konvolve edilmesiyle öznitelik çıkarım işlemi gerçekleştirilmiş ve filtre çıkışındaki gürültünün etkisini önlemek için çevrelenmiş piksellerin yoğunlukları ortalama olarak alınmıştır. Her bir fazör için iki data biti üreten her bir filtreyle faz dört seviyeye quantize edilmiştir. Son olarak eşleştirme için bitsel karşılaştırmalar gerekli olduğundan Hamming uzaklığı kullanılmıştır.Çalışma sürecinde filtrelemeler, görüntü işleme gibi kullanılan süreçlerin tümü MATLAB ortamında gerçekleştirilmiştir. Bu tez çalışmasında gerçekleştirilen sistemin performansını test etmek için ROC analizi yapılmıştır. Analiz sonuçlarına göre, geliştirilen sistem kimlik tanımlama hassasiyeti ile tatmin edici sonuç sağlamıştır.
IDENTITY RECOGNITION TROUGH ANALYSIS OF IRIS IMAGES Biometric systems which can response to additional security needs in peresent technology, is an image matching systems which uses original physcal and behavioral properties of which people their biological products have. Biometric systems use only properties of person, therefore remembering password or carrying ID card is not required. These systems record the personal information and make up a database, and then used for every security inquiry. In the scope of this thesis studies have been implemented for iris recognation systems based on iris which is accepted as one of the most reliable biometric characteristics due to reasons like its highes discrimination ratio between human beings, being a well-protected internal organ, being impossible to imitate, low genetical likeness and no contact while capturing the image. Canny edge detection algorithm and Hough transform have been used in order to obtain iris recognation from the images in the database. Iris image extracted from eye. Images have been trcensformed from polar form to rectangular form. In order to obtain the iris zone from the pictures on the data base used on the system, Canny edge detection algorithm and Hough transform were used. The iris picture which was cut at received eye pictures was transformed from circular form into rectangular form. Normalized iris design was convoluted by Gabor ripples and so, self-quality deduction procedure was obtained and, in order to prevent the effect of the noise at the filter outlet, the density of the surrounded pixels was taken on an average. The phase was encoded to four levels by each filter generating two data bits for each phaser. And finally, Hamming distance was used since bit comparisons were necessary for matching. During the study, all the processes such as filtrations, picture processing etc were done at MATLAB environment. ROC analysis has been carried out in order to test the performance of the thesis study. According to the analysis reports, the developed system has been provided satisfying results through ID recognition precision.
IDENTITY RECOGNITION TROUGH ANALYSIS OF IRIS IMAGES Biometric systems which can response to additional security needs in peresent technology, is an image matching systems which uses original physcal and behavioral properties of which people their biological products have. Biometric systems use only properties of person, therefore remembering password or carrying ID card is not required. These systems record the personal information and make up a database, and then used for every security inquiry. In the scope of this thesis studies have been implemented for iris recognation systems based on iris which is accepted as one of the most reliable biometric characteristics due to reasons like its highes discrimination ratio between human beings, being a well-protected internal organ, being impossible to imitate, low genetical likeness and no contact while capturing the image. Canny edge detection algorithm and Hough transform have been used in order to obtain iris recognation from the images in the database. Iris image extracted from eye. Images have been trcensformed from polar form to rectangular form. In order to obtain the iris zone from the pictures on the data base used on the system, Canny edge detection algorithm and Hough transform were used. The iris picture which was cut at received eye pictures was transformed from circular form into rectangular form. Normalized iris design was convoluted by Gabor ripples and so, self-quality deduction procedure was obtained and, in order to prevent the effect of the noise at the filter outlet, the density of the surrounded pixels was taken on an average. The phase was encoded to four levels by each filter generating two data bits for each phaser. And finally, Hamming distance was used since bit comparisons were necessary for matching. During the study, all the processes such as filtrations, picture processing etc were done at MATLAB environment. ROC analysis has been carried out in order to test the performance of the thesis study. According to the analysis reports, the developed system has been provided satisfying results through ID recognition precision.