| Biological recognition in recent years gradually brought to the attention of the people and,due to the information security has increasingly become the focus of attention,how to more effectively protect the individual enterprise even is the property of the state of information security becomes an urgent topic.As a unique attribute of individual,biological characteristics have natural advantages.The iris recognition as a more accurate biometric method is also the object of many scholars' research.In this paper,the key problems of iris recognition stage are used as the starting point.(1)Creating a laboratory in the fourth generation of the iris image library,the image will be divided into 20 classes,including clear iris image,the iris images,eye eyelash shade image,eyelids cover image,blurred image,etc.There are 70 images in each of these categories,and 1,400 images.The image library has been collected and used for nearly a year and has achieved good results.It has provided assistance to the team's research and study,and has become the database source of the attendance system.(2)An improved quality evaluation algorithm is proposed.Mainly adopts distributed,multiple indicators into fusion classification,the method of evaluation can be divided into coarse assessment and evaluation of two big stage,in the evaluation of coarse,including images overall sharpness detection and open degree,in the overall sharpness detection,the fuzzy image directly deleted,to avoid disturbance to the subsequent operations.In the detection of open eyes,the image of the non-conforming image is screened by the method of the pupil boundary location,and the image is discarded directly.Through rough evaluation,some of the collection effects can be poor,and the subsequent identification process can bring unfavorable conditions for image elimination.Through rough evaluation,after woke up to image evaluation,including the iris sharpness detection,pupil detection and obstructions deviation degree,evaluation index is obtained by the three after the operation,then the index data based on SMO algorithm of support vector machine(SVM)classification,finally the image equalization processing,to get the final result.The main innovation points of this algorithm are focused on the planning of the process,and the pupil deviation detection is added,and after the classification,the qualified image is balanced to facilitate the follow-up operation.(3)An improved localization algorithm is proposed.At first when the positioning of the internal circle pupil after iris image quality evaluation of qualified morphological open and close operation,the dark details in the iris image point to weaken,then the image binarization operation,get the corresponding binary image,convenient follow-up operations.In this way,the edge detection of binary images can be detected by the difference of the grayscale of the pupil and the obvious iris,which can detect the edge points near the pupil.Finally,Hough round test is carried out to obtain the final inner circle boundary.Interested in and outside the boundary location is the first selected area,concluded that iris texture exist in the area,then use Canny operator for edge detection,in the territory through some edge point of attachment of the median line get external boundary's center and radius,then use to draw out the Hough circular boundary,so as to locate the iris of the internal and external boundary.The innovation points of the localization algorithm are mainly focused on two points,and the open and closed operation is used in the internal boundary positioning to improve the recognition accuracy.The concept of interest area is used in the outer boundary positioning,which reduces the positioning range and makes the search operation speed increase.Finally,through the comparison test verification,the method proposed in this paper is effective,which can improve the recognition performance of the system and has certain practicability. |
PDF Full Text Request