Study On Algorithm Of Iris Recognition

Based on biometric identification technology is a automatic identification technonglogy, which is based on human physiological and behavioral characteristics to identify or verify a person's life. The special nature of the iris is to make use of iris identification of an extremely high accuracy rate. Iris recognition technology is more and more attended in the research and application, and widely considered to be the most promising biometric technology.In this paper, there are a comparative study of several typical iris recognition method, a focus on algorithms such as iris positioning algorithm, feature extraction algorithm and iris matching algorithm, some improvement of a iris positioning algorithm and feature extraction algorithms, and some ideas for further study. The main content and research results are as follows:Comparative studied of algorithms such as Daugman iris location algorithm,Wildes iris location algorithm and the iris location algorithm of the CAS institute of automation Wang Yun hong, Tan Tieniu and so on, analysised their respective advantages and disadvantages and improved Canny operator of the Hough transform location algorithm. Improved algorithm using Canny operator to detect edge, using threshold segmentation and gray projection to locate inside edge and using the method that search white pixels to locate outside edge. The experimentation proof that Improved algorithm not only enhance the effect of location, but also lowered the time complexity.Using Daugman rubber model to unify the iris image and using gray histogram equalization to enhance the iris image ,which is unified.Comparative studied of encoding methods such as the encoding method of Daugman system, the encoding method which is based on two-dimensional Log Gabor transform and the encoding method of Boles, analysised their respective advantages and disadvantages and improved the algorithm that detects the zero-crossing with a wavelet. Improved algorithm detects the zero-crossing of zero-crossing with a wavelet. Improved algorithm detects the zero-crossing of unified image at along the radius and arc, extractes the feature signature of zero-crossing to encode and improves the performance of translation and rotation invariance. The experimentation, that measure the similarity, proof that Improved algorithm improved the rate of recognition, which is more than 97%.The improved algorithms and ideas enriched the iris recognition research, has atheoretical significance, nevertheless, there are still shortcomings in some areas andneeds to further study.