ALGORITHMS OF SEPARATION OF TEXTURAL MARKS OF THE IRRITANT EYE
O.G. Oksiyuk, Andriy Fesenko, V.I. Vyelkova, Valeriyan Shvets · Collection of scientific works of the Military Institute of Kyiv National Taras Shevchenko University · 2019
Undoubtedly, human authentication is an urgent task, a practical solution that employs thousands and millions of people around the world. The tasks of authentication and human identification are now solved with the help of automatic biometric systems, constituting one of the new fields of applied mathematics, biometric identification. From the point of view of reliability, the most effective methods of identification and authentication today are biometrics, which allow to solve the problems of losing passwords and personal identifiers. Among biometric technologies, one of the most promising is biometrics with the use of the iris, which has a specific structure and contains a lot of textural information. Spatial structures observed in the iris are unique to each individual, and individual differences appear in the process of anatomical development. The limiting factor for the proliferation of Iris systems has always been their high cost, but ongoing research and development will reduce costs, and expanding the scope will allow authentication technology for Iris to occupy a prominent segment in the access control market. The paper analyzes the disadvantages of iris processing using the Gabor mathematical apparatus used by Dr. John Daugman and offers an alternative method of extracting informative features from the image of the iris, based on the use of a DoG filter. A feature of the DoG filter is that its response changes the mark in areas of the image where there is a difference in brightness. In homogeneous areas of the image, the response of the filter is zero, but there are almost no such areas in the image of the iris. The advantage of using a DoG filter is that only the Cartesian coordinate space, which is natural for image processing, is used to calculate it, and the features obtained provide better class separation than features based on Gabor filters.