The Research And Realization Of Iris Location And Recognition

In biometric identification, fingerprint is most commonly used method and is a kind ofrelatively mature identification method. But its limitation is that when fingerprint is not clear,the identification accuracy will descend, especially when finger is injured or lost, it willdisenable identification. To remedy this limitation, a multi-biometric identification is adopted.It means that, when one biometric identification method becomes invalid, the other methodcould still ensure the normal operation of system. Iris of man’s eye could be widely usedsince it is a stable, unique and un-contacted method for identifying object. This dissertationfocuses on improving the matching accuracy and speed of the identification method based onexisting algorithms, and establishing quality evaluation system for iris image. The maincontents are as follows.In the aspect of improving iris detection accuracy,(1) to many cases of elliptic pupil,algorithm improvement is made to this situation by calculating the intersection pointsbetween elliptic pupil and iris, and radial line which is from the center of pupil and arrayed byevery certain angle. Then small area of elliptic pupil could be obtained.(2) to the cases ofdeformed image, the thought of partial difference is adopted to pursue the greatly changedposition. Previously scanning for iris boundary only was along with vertical axis. When irisimage is deformed, such scanning image is more accurate. By adopting the thought of partialdifference, starting from the position, certain angle between the left side and right side ofvertical axis, with different scanning radius, the obtained position is the iris boundary inwhich gray intensity is changed most greatly.(3) to the image noise treatment, de-blurringtool and Gauss filter are used. Image blurring is often caused by movable camera equipment. This kind of blurring may be an image interlace. By using de-blurring tool, the clearness isimproved. High-frequency noise is reduced by using a3×3Gauss filter, and matchingaccuracy has been improved a little.In the aspect of establishing quality evaluation system for iris image, in order to realizequality evaluation for iris image automatically and in real time, one should make analysis andassessment for each acquired image. The quality of image must first meet lowest assessmentpoints. On the basis of passing lowest assessment standard, then each single assessment itemis judged. Only images meeting minimum requirements could be put in the next step forfeature extraction. When the integrated iris quality assessment is considered, pupil radius/irisradius, iris radius-pupil radius, iris visibility, and sum of gray intensity difference of adjacentsmall area are considered as assessment elements. The algorithm is simple and easy to beimplemented.In the aspect of manual pre-judging tool for iris image, the proposed system has thefunction of locating pupil and iris and eyelid intuitively and in real time. It has also thefunction of feature extraction and matching. The modules used are conformed to the ones indeveloped program. Thus it could dynamically evaluate the correctness for each module iniris system.The main contributions of this work are as follows:(1) To many cases of elliptic pupil, by adopting the one order differential, scanning withan arc in horizontal axis and vertical axis separately, the obtained position is the iris boundaryin which gray intensity is changed most greatly.(2) To the cases of deformed image, the thought of partial difference is adopted to pursuethe greatly changed position by scaning horizontal and vertical axis.(3) A quality evaluation system for iris image is proposed, which evaluates iris qualitybefore locating pupil and iris, and recognition accuracy is improved in iris recognitionsystem.