The Research Of Iris Recognition Algorithms Based On Fractal Geometry

With the rapid development of global economy and the advancement of science and technology, the demands for security are becoming higher and higher. The conventional personal identity authentication can't meet the need of information security. The way of recognizing identity by physical characteristics of human body, such as fingerprint, iris and voice, has been put forward. Due to its non-invasiveness, high reliability and other good features, Iris recognition technology has become the research hotspots of Biometric Identification Technology.Iris recognition system is mainly composed of iris image acquiring, Iris localization, image preprocessing, feature extraction as well as matching. This article focuses on iris recognition algorithm which is mainly used in preprocessing, feature extraction and mode matching.This paper locating iris area based on gray characteristics of iris. First, a point within the pupil is found, and then from this point, we search three points on the edge of a circle which were not on the same line, in the last, we can locate the inner and outer boundary with the principle that three points which are not on the same line can define a circle.Because the gray changing is not obvious when locates outer boundary, we detect four outer boundary points to locate through two times and chooses a better result as the locating result. This improvement can reduce measurement error and bring a higher accuracy and precision of location than one time to locate.This paper proposes a novel iris feature extraction algorithm using combination of blanket dimension and lacunarity. Because Human iris texture is characterized by fractal geometry due to its rich self-similarity and abundant variation, vertically expanded blanket dimension is employed to represent iris texture variation and radical pattern at different resolution levels. Lacunarity is introduced to extract feature from two irises that have different texture and fractal patterns but have the same fractal dimension value. The combination of blanket dimension and lacunarity in iris feature extraction can embody the minute change of texture information comprehensively, and improve the capacity of iris classification.Normalized correlation classifier is used for pattern match and cyclic shift algorithm is introduced to eliminate the difference resulting due to the iris rotation in this paper. 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 7.0 with the CASIA-IrisV3-Interval iris database, proving that the algorithm developed in this paper is effective in improving the recognition accuracy and speed.At last we make the conclusion of this paper. At the same time, some expectations to research are predicted.