Study On Algorithm Of Iris Recognition

Biometrics is a new identification technology, because of its high safety and security, now it is concerned by more and more scientists. As one of the biometrics, iris recognition technology has the characteristics of high security, uniqueness, stability, availability of acquisition and non-invasiveness, and is regard as a technology of great research value and wide application prospects.In this paper, the background of biometrics and several existing biometrics are introduced firstly, and their characteristics are compared; then the iris recognition technology and the iris recognition system are discussed in details; finally, the algorithm of iris recognition is studied, and an algorithm of the iris local texture feature extraction based on Gabor filter is proposed, and the relevant experiments are carried out in the CASIA iris database. Iris recognition algorithm mainly comprises image positioning, normalization, feature extraction and match. This paper makes studies on all these parts, and the main jobs are as follows:1. A fast iris location algorithm based on minimal variance searching center is studied. Firstly, the region growing method is applied in the algorithm to eliminate the interference of bright spots; then based on the gray distribution features of iris, projection method is applied to get the circle parameters of iris'inner edge; lastly, the circle parameters of iris'outer edge is obtained by applying the Canny operator, the prior knowledge, the designed algorithm using the circle parameters of inner edge.2. The iris image is normalized by applying Daugman's rubber paper model method, and the rectangular image of iris with the feature of translation and scale invariance is gained. Then the normalized iris image is enhanced to improve the contrast of the iris texture. Finally, the effective area of iris is determined to reduce the impact of eyelids and eyelashes on iris texture.3. An iris feature extraction algorithm which extracts the local texture feature based on the Gabor filter is studied. The algorithm makes the absolute value of the output image's filter coefficients as a weight, then operates weight average on all the pixels of the filtered image and gets the feature points of iris. In the process, an optimal weighting method is proposed. It makes the bigger filter coefficients greater contribution to feature points and the smaller filter coefficients less. Finally, the Euclidean distance classifier is designed to classify the feature points. The algorithms designed in this paper are applied to the sub CASIA iris database which is provided by Institute of Automation, Chinese Academy of Sciences, and the results show that the algorithm have a higher recognition rate which is up to 99.77%, and it is relative stable for the rotation of iris.