The Research Of Iris Location Problem

In age of electronic information we need a more convenient and effective identity verification technology. Biological recognition technology makes use of the inherent physiological or behavioral characteristics of human beings to identify individuals. Due to the unique, stable and live physiological properties of the iris and non-invasive to users, most stable and reliable for identification in practical applications of iris-based system, iris recognition is receiving extensive attention and becoming an active topic in biological recognition technology.The iris preprocessing is a very important part of iris recognition system. The quality of the iris image preprocessing directly determine speed and accuracy of the entire iris recognition system . The definition of the iris image preprocessing in this paper is the treatment of original input iris image sequence until we can find the iris feature in the original image.After the iris image preprocessing ,we can straight extraction feature to match. The main work of this paper is the research of iris location algorithm .In this paper,first we introduct Daugman's circumference match method and Wildes' two-step method, the optimal gradient method, Zhongkeyuan Tan Tieniu's method, H. Proenca's based on the FCM iris segmentation method, then we analyse the advantages and disadvantages of these algorithms. H. Proenca's based on the FCM iris segmentation methods had a good performance with the UBIRIS database , but the iris pixel whose gray has relatively small difference with the sclera pixel will be easily clustered wrong, especially the iris part near the outer edge of ((?),(?)), ((?),(?)). After analysis,we believe that the method H. Proenca used,with the spatial information (x, y) clustered ,the cluster center of the coordinates information appeared to be only one point and only a few pixels around the center ,most of the pixel is far away from the center of coordinates information ,it result in the gray information played a decisive role in the cluster process. So it is not difficult to understand why some points in the iris can be easily be wrong clustered . Daugman's method has good robustness and accuracy, even if the image is not clear can be good location.But Daugman's method searched from the center along the direction of the radius, is likely to fall into the local lowest. The Daugman's method will cost lots of time if it search the whole image. The key to solve these problem is how fast had the range of the iris' outer radius.By the above considerations, we have the radius information replace the coordinate information . After clustering, the radius information center in the images is showed a round and most of pixels' radius information surround its radius information center. Of course radius information we used accord with the iris' circular propertyTo locate the pupil, we use gray projection methods to decide the inter circle of the iris.For every pixel of the image ,then we use curve fitting method to locate inside ciecle.we have the distance to the center of inter circle ,the distance we have is used as radius information of the pixel.With the radius information and the gray information we use IKFCM method to have the range of outer radius .Based on the Daugman's operator, we construct a operator:with the radius between(r-10,r+5),the center between (xx±2x±4,yy±3y±6),we search the (x,y,r) to make the operator we constructed has biggest value ,the (x,y,r)we make is the parameter of the outer circle.We have the experiment in the CASIA1.0 Datebase,we find that the method we use can finely cluster the image ,the radius of outer circle first located minus the real radius is in the range [-5,10].The iris image preprocessing relate with specific iris image acquisition equipment and the quality of image.With different iris recognition system,we need a corresponding specific preprocess methods, there are lots of problem will be resolved .These will need our further study.