Research On Iris Identification Method Based On Texture Feature

With the rapid development of modern information technology, the biometric identification technology has become an important research topic in the field of information security. In recent years, the iris recognition technology is given great attention in the area of personal identification technology, because of iris has the feature of uniqueness, high stability, ant-counterfeit, which possessed by the individual. The Experts and scholars around the world have made great contribution to the development of iris recognition in theory and application.The thesis is completed under the Science and Technology Project of Sichuan, "Research on physiological information collection technology based on Internet of things". The main research contents of this topic:(1) The establishment of the iris image databases includes the standard database and the test database.(2) The iris image preprocessing, this mainly includes boundary localization, normalization and enhancement of iris image.(3) The iris texture feature extraction and encoding that quantize texture feature information.(4) The iris texture feature matching that is compares the collected iris images with images in the database, and then realizes the identity.The research routes and methods:(1) For validating the performance of the proposed methods two databases of iris images are used, one is Iris Image of Version 1.0 from the Institute of Automation of the Chinese Academy of Sciences, and the other is iris images of laboratory.(2) In the process of image preprocessing, several common iris localization methods are introduced. And an improved iris localization algorithm based on Hough transform is proposed, which has better accuracy, speed and robustness. In the stage of iris image normalization and enhancement, iris region from the Cartesian coordinates is remapped to the polar coordinates through building a flexible model, thus the iris image is normalized. The method of histogram equalization is used in the normalized image so that the texture information of iris is highlighted. Enough texture information area of iris is chose as effective region from the normalized image.(3) In the section of iris texture feature extraction and encoding, the theories of algorithms are introduced and analyzed, which originated by Daugman, Boles and Lim. Iris texture feature is extracted by use of a zero-crossing operator and encoded by figuring up the numbers of the binary-conversion for the signs of results. The proposed algorithm not only has high computing speed but also overcome the problem of image rotation.(4) In the course of iris feature matching, two improved algorithms are proposed by presenting and analyzing some widely used algorithms, one is improved classification algorithm based on weighted Euclidean distance and the other is improved matching algorithm depend on Hamming distance. Both of them have good performance in image de-noise and overcoming the problem of image rotation.The research results and conclusions:Through in-depth researching for the key technology of iris recognition, corresponding improved methods are put forward and a series of experiments are conducted.The main innovation points:(1) the rules of horizontal edge selection are used to remove the non-horizontal edge points and the coupling relationship between the internal boundary and the external boundary are employed to determine the parameters of boundaries of the edge-selected iris image, make average positioning time is 0.152 s and positioning accuracy is 98.4%; (2) the zero crossing detection operator and the secondary encoding method reduce the complexity of the calculation and overcome the influence of image rotation; (3) the two improved algorithms of iris feature matching are effective and the weighted Euclidean distance classification algorithm obtains the accuracy rate of 96.04% that better than others. The results show that the validity and practicability of the proposed methods.