The Research On Iris Recognition Based On Snake Model And Independent Component Analysis

With the information age arrival and network technology broad application, people's requirement for security is higher and higher.The recognition technology based on biology characteristics of human body got rapid development and application. As an important feature of personal identification, iris has some advantages such as uniqueness, stability and no-infringing etc. So iris recognition is considered one of the most accurate and reliable biometrics at present.In this thesis, the traits of iris as a biology characteristic were elaborated from the iris physiology, the unique superiority about the iris to be used in the identity recognition was summarized. The basic system framing based on the iris identity recognition was introduced: the iris image preprocessing, the feature extraction, the matching and the recognition. The several kinds of quite mature iris recognition systems at present were analyzed and compared, their respective advantages and disadvantages were elaborated. The previous work of other researchers was summarized, the research on key technologies of iris recognition was processed, and several improved new methods were proposed as following:Firstly, in the iris image preprocessing phase, a new iris localization algorithm based on Snake model was proposed: the Canny operator was selected to locate the inner edge of iris, the Snake model was used to lock the outer edge, and the polar coordinate conversion was made use of carrying on normalized processing of the iris image. In order to reduce the influence of unsymmetrical illumination, an iris texture enhancement method based on the wavelet was presented. Secondly, in the process of feature extraction, a new iris recognition algorithm based on Independent Component Analysis (ICA) was proposed to extract iris feature. Finally, Support Vector Machine (SVM) was used to carry on the pattern matching, because SVM performs pattern recognition for two-class problems by determining the optimal linear decision hyperplane based on the concept of structural risk minimization.Experimental results show that the method proposed in this thesis has satisfying performance both in speed and precision, also has good robustness in iris location. Compared with other methods, the size of iris code and the processing time of the feature extraction are significantly reduced.