Research Of Iris Recognition Technology Base On SIFT

As a new kind of identification technology,biometrics is widely used in the areas of defense,e-commerce,finance and education.Among the many biological features,the iris feature has the characteristics of high stability,high anti-counterfeiting and uniqueness.Therefore,it has been the focus of research in the field of biometrics.In the real application scenario,the problems such as the incorrect posture and the skewed head make the images rotated and deformed.The accuracy of recognition using the classic iris recognition algorithm is not high.Therefore,it is of great practical value to study the identification technology of rotating and deforming iris images.The main research work of this thesis includes the following three aspects:1)In order to solve the problem of inaccurate rotation iris image localization,an S-R-F iris localization algorithm is proposed in this thesis.Firstly,locating the inner edge of the iris,using the histogram equalization method to find the appropriate threshold,and then using the threshold to finish the iris image binarization.Secondly,using morphological open operation to remove noise,and then searching the pupil to determine the location and center of the circle radius.When the outer edge is located,the outer edge is not a regular circle due to the rotation and deformation problems.Therefore,the outer edge is calculated by the method of Canny edge detection and RANSAC algorithm when fitting the ellipse.Because of adopting RANSAC algorithm,an effective outer edge point set can be obtained,thereby improving the positioning accuracy of the outer edge.2)In order to solve the problem of feature extraction and matching inefficiency of rotating iris images,a H-S-P iris feature extraction algorithm is proposed in this thesis.Firstly,the Harris algorithm is used to detect the corners,and then the SIFT operator is used to characterize the features.Finally,the SIFT features of the points are obtained for PCA dimensionality reduction.The algorithm avoids the operation of the existing algorithm when it is matched to the cyclic shift,which can effectively improve the matching efficiency.At the same time,this thesis uses the neighborhood distance algorithm to match the features,which avoids the open-root operation of the Euclidean distance and improves the running efficiency of the algorithm.3)In order to verify the effectiveness of the proposed algorithm,this thesis takes the image of CASIA iris database as experimental data and the existing iris recognition algorithm as contrast experiment.The proposed algorithm is compared with the existing algorithms by four indicators: rejection rate,false positive rate,ROC curve and correct recognition rate.Experimental results show that the proposed algorithm is superior to the existing algorithms in accuracy.