Biometric algorithms for iris recognition

MASARYKOVA UNIVERZITA FAKULTA INFORMATIKY Biometric algorithms for iris recognition Bachelor’s degree thesis Petr Baštan Brno, 2013 Declaration H...
0 downloads 1 Views 223KB Size
MASARYKOVA UNIVERZITA FAKULTA INFORMATIKY

Biometric algorithms for iris recognition Bachelor’s degree thesis Petr Baštan

Brno, 2013

Declaration

Hereby I declare that this paper is my original authorial work, which I have worked out by my own. All sources, references and literature used or excerpted during elaboration of this work are properly cited and listed in the complete reference to the due source.

Petr Baštan

Supervisor: Ing. Mgr. Zdeněk Říha, Ph.D.

Acknowledgement I would like to thank my supervisor, Ing. Mgr. Zdeněk Říha, Ph.D., for his continuing support, encouragement and supervision during my work on the thesis. Special thanks goes to my family for the support during the whole work.

Abstract The thesis is devoted to biometric recognition techniques based on iris images. Iris recognition with its complex random patterns should theoretically be, among other biometric systems, the most accurate method for the human identification. In the text of the thesis are compared known algorithms for iris recognition with taking into account the accuracy and the speed of the processing of the iris images. The thesis also covers the basics of the biometrics systems and medical background for the eye and the iris itself. The practical part of the work implements one of the algorithms and evaluates its performance on open databases with iris images with direct comparison to the results of the previously compared algorithms. The results are showing how the false and positive rejections and acceptations are affected by chosen algorithm and by implementation itself while providing recommendations for future development for other systems.

Keywords Biometrics, iris recognition, authentication, verification, FAR, FRR, DET, FNMR, FMR

1 Introduction

1.1 Biometric systems

Biometric systems enable to automatically authenticate a person based on traits or measurable characteristics. They are being used in many facilities such as government buildings, airports and in general areas with a need for a secure authentication systems. Examples of following biometric systems are being distinct: systems based on fingerprints, voice, hand geometry and handwriting, the retina and the iris. Biometric identifiers are unique to individuals to a certain level hence they can be recognised as more reliable while verifying the identity compared to more traditional token or knowledge-based methods. The identifiers can be also categorized as physiological and behavioural characteristics of the person.[1] In the mentioned systems first of all the sample of a biometric identifier is being captured, then while using a mathematical function/apparatus the sample is being computed into a template. The template serves to provide a discriminating representation. Biometric systems allow to perform an enrolment mode for collecting templates and an identification mode, where a sample taken later on is being compared with the templates in the database until the system either finds a matching template or rejects the sample and the subject becomes unidentified.[1] The variability among different instances of a given class is one of the most important features within biometrics and vital for the reliability and the robustness. Most of the physiological and behavioural characteristics might change over time, such as the tone and the pace of a voice, also the face expressions or the signature shape are changing significantly during certain periods of life. The variability is also in many cases affected by the environment where the samples of the traits and characteristics are being taken.[1] The illumination, pose or viewing angle of the face, speaking while under stressful situation or in different age might result either in a false acceptance or a false rejection of a subject, in both cases it is not a desired completion of the whole process. The methods for authenticating a person can differ in each stage of the process – in securing the correct liveness detection, capturing the sample, creating the template or even in the matching criteria. The performance and results of the biometric methods are in the end affected by high number of factors.

1.2 Iris The iris is starting to be shaped within the third and the twelfth month of embryonic life. The complex pattern of the surface of the iris contains distinctive features (e.g. furrows, crypts, corona, freckles). The patterns themselves are genetically independent and unique in such way that there are not even two irises which might be having exactly the same patterns due to the epigenetic nature of their patterns and will remain uncorrelated.[2] The function of the iris is to control the amount of light entering through the pupil by using the sphincter and adjusting the size of the pupil with the dilator muscles. The iris consists of a the epithelium layer and stromal layers. Those layers contain pigmentation cells, blood vessels and iris muscles. The visible surface of the iris contains two zones - an outer ciliary zone and an inner pupillary zone which are divided divided by the collarette.[3]. In addition, the iris is an internal organ of the eye therefore it is protected well enough from the outer environment and is not a subject to change over time. Those features provide iris recognition among other biometric technologies bigger reliability and makes it suitable alternative to other methods.

1.3 Illnesses of iris As in any other method being used within biometric systems any serious injury or disease of an individual has to be taken into account while designing this system rather than algorithms themselves. There might be a difficulty obtaining the necessary characteristics once the individual loses the access to the personal funds or the comparing samples after initial enrolment with later ones, which are significantly affected by certain circumstances. Even while being highly protected human organ, iris still can suffer from various acute diseases.

1.3.1 Uveitis Uveitis, i.e. inflammation of the uvea, is a serious ophthalmology disease with ability to change the shape of the iris in a human eye. In a study conducted by Tariq Mehmood Aslam, Shi Zhuan Tan and Baljean Dhillon, patients with uveitis had been involved in testing of applied Daugman's algorithm. The recognition failed in 20,83 percent of cases, other pathology forms have not significantly affected the whole process although not all of them have been conducted into the study due to limited samples. The general cause of the iris recognition failure with subjects suffering with this disease has not been determined, although pharmacological pupil dilation, pathological pupil distortion and abnormal variations of a pupil size were suggested as the most probable cause. [4]

1.3.2 Aniridia Aniridia is referred to as a complete or partial absence of iris with a possibility to include both of the individual's eyes. This congenital disorder may occur in an association with other abnormal conditions of a human eye and the iris recognition will fail under such circumstances. 1 in 50,000 to 100,000 newborn children worldwide is being affected by this disease, therefore the potential of having a subject with similar disorder in a database with enough samples is considerable.[5]

Bibliography

[1] KOLEKTIV. Biometrie a identita člověka: ve forenzních a komerčních aplikacích. Praha: Grada, 2008. ISBN 978-80-247-6392-7. [2] R. Wildes. Iris recognition: an emerging biometric technology. Proceedings of the IEEE, Vol. 85, No. 9, 1997. [3] Encyclopaedia Britannica: Sensory Reception: Human Vision: Structure and function of the Human Eye. Michigan: Encyclopaedia Britannica, inc, 1987. ISBN 9780852294437.

[4] Aniridia [online]. 2009, 2013 [cit. 2013-05-18]. Available on: http://ghr.nlm.nih.gov/condition/aniridia [5] ASLAM, Tariq Mehmood, Shi Zhuan TAN a Baljean DHILLON. Iris recognition in the presence of ocular disease. In: s. 2-4. DOI: 10.1098. Available on: http://rsif.royalsocietypublishing.org

Suggest Documents