The Research Of Several Algorithms On Iris Recognition System

It is a highly interacted society in modern times, in which the issue of identity authentication involves into every aspect of our daily life. Through identifying articles such as the key and the id card, or through identifiers like password, code or argot, which are the traditional methods of authentication but unable to meet the current needs. So it is extremely urgent to research on a new technique. In this case, biometric authentication is convenient and reliable way to test the user through biometric recognition of intrinsic biological features and the daily-life habits. Therefore, the biometric way is promising in this field. Then iris is definitely going to be one of the main techniques of identity authentication because it is with a fast speed and a high stability but without any invasion to the users.Iris recognition system consists of five parts: image acquisition, image quality evaluating and counterfeiting, image-preprocessing, image feature extracting, and finally feature matching. Based on the predecessors achievements, this paper tries to do research on the algorithm in the authentication process and to make improvement to combine the practical need. My main tasks are as follows:from the aspect of rough location, I have studied on the common algorithm of coarse location, and then improve the performance of gray geometric projection approach by combining mean filter and histogram statistic property. Hence it can adjust the size of the window by self-adaption. Then a coarse location algorithm based on line segment markers is put forward, which is united with geometric projection and to achieve relatively fast and exact location of the pupil.from the aspect of accurate location, In this thesis, I carefully study of based on Hough Transform for circle detection algorithm. For the problems of its large calculated amount and long consuming time. This paper presents an improved based on Hough Transform for circle detection algorithm. Through wiping off some distraction and reducing the calculated amount, so the searched volume of the Transform parameters can be narrowed. Hough transform will reduces the computation and improve its real-time.from the aspect of feature abstraction and matching, through carefully analysis of iris texture information in different frequency bands the contribution of identification, this paper presents a cycling coding method of local airspace Zero-Crossing detection. this new way is capable of solving the slump of recognition rate under uneven illumination or in the noise jamming conditions, and in those circumstances; the images would be affected badly. Finally, it takes advantages of SVM to solve the small sample, nonlinear and high dimensional pattern recognition, optimizing of iris feature matching.