The User Is Less With The Iris Identification Methods Of The Research

Non-cooperative iris recognition, which has several advantages, such as better adaption to marketing and efficiency, has become the primary development trend of iris recognition technology. In iris images which captured in the non-idealconditions, iris is always incomplete, deforms and blurred which as a result of the occlusion, moving, reflection and other different interference and noise. In such case, it’s challenging to get accurate recognition performance.First of all, a method of iris location for CASIA-IrisV4is described, whichcalculates the binary image to keep reflection region near pupil, determine approximate position of pupil by the horizontal and longitudinal projection of binaryimage, finally get the fitting iris outer and inner boundary circle by using Cannyedge detection and Hough transform within the qualified range. In iris registration process, inverse composite algorithm based on sub-image is adapted. This paper propose two iris feature representation algorithms: One is based on Gabor filters via extracting phase information. The other is WCPH(Weighted Co-occurrencePhase Histogram). The former combines floating search method, and select Gabor filter set with the best classification performance. DI is adopted to measure the classification performance. The latter method organizes WCPH vectors according to space position, and use Bhattacharyya distance to measure the similarity.Experiment is conducted on UBIRIS.v2and CASIA-IrisV4iris database, andROC curves and DI(Decidability Index) is used to evaluate iris recognition performance. In2010, we participated in NICE.II(Noisy Iris Challenge Evaluation), ranked NO.7among67teams from30countries and regions.