Research On Iris Recognition Algorithm

With the development of network and communication technology, and the coming of information age, the conventional identity authentication can't meet the need of information security, so the method based on biometrics emerges as a new identity authentication. Iris recognition technology is one of the biometric identification technologies, thanks to the characteristics of iris such as uniqueness and stability, the error rate of iris recgnition technology is the lowest among the several common biometric identification technologies, and iris recogniton technology is also most approbated. Iris recognition technology has become the research focus, and both the recognition algorithm and application research have wide development prospects.Iris recognition system is made of four parts: image capturing module, iris preprocessing module, iris feature extracting module and pattern matching module. Image capture means the capture of original eye image, in this paper we use the CASIA iris database supported by Chinese Academy of Sciences as the original images; Iris preprocession includes three steps, namely, locating the inner boundary of iris with excluding the noise interference such as eyelids and eyelash at the same time, normalizing the preprocessed iris image, and image enhancing; iris feature extraction means extracting useful features from a mass of iris features and encoding them to form specific iris template; pattern match is comparing the objective iris encoding with the templates in the database to calculate the similarity, and then obtainding the match result.In this paper, we firstly introduce the knowledge of iris recognition, including the physiological characteristics of iris, the theoretical basis of iris recognition technology, and the structure of iris recognition system. Then we research the iris recognition algorithm, the main work is shown as follows:Image proprecossion: On the base of analysing several classical locating algorithms, the paper developes a new locating algorithm which basic idea is using threshold segmentation and Point Hough Tranform to locate the inner boundary of the iris image firstly, and then using Canny operator and Circle Hough Transform to locate the external boundary; normalizing the image by transforming cartesian coordinates to polar coordinates; enhancing the image with histogram equalization to enhance the contrast of the iris image.Iris feature extraction: this introduces common texture analysis methods, encoding methods and an effective classification and reduction mathematical tool rough set theory. On base of the frontal introduction, a new algorithm is developt which reducting the iris features with rough set theory to obtain the most effective information and encoding with 1D Log-Gabor filter to form the specific iris template.Pattern match: this paper does pattern match with Hamming distance classifier, on which there is some improvement, namely introducing cyclic shift algorithm to eliminate the difference resulting from rotation. Research on the optimal setting of key parameters in the algorithm is also done to improve the matching rate.All the above algorithms are simulated on the platform of Matlab 6.5 with the CASIA V1.0 as the original images which includes 756 eye images from 108 different eyes. The results are verified in both authentication mode and indentification mode, proving that the algorithm developt in this paper is effective in improving the recognition accuracy and speed.In the end of the paper, summarizition of the subject is done and there is a prospect of the following work.