Iris Recognition System Based On CS-LBP Operator

Iris recognition system has matured after several years research and development ofrevolution, but in many aspects is not ideal. It is hard to find the balance between the accuracyand the speed. Improving the user experience and strengthening the recognition pertinenceand integrity is the purpose of this paper, optimize the existing recognition technology toestablish a rapid, effective and friendly real-time iris recognition system.Iris recognition algorithm includes iris image quality assessment, iris imagepre-processing and iris feature extraction and pattern matching, the main work and result asfollows:In the iris image quality assessment link, this paper propose a comprehensive andstep-by-step assessment method. Firstly introducing the concept of “Interesting Area” toavoid to scan the whole iris image; secondly using the Gauss Laplace operator, dimensionssmaller and detection range wider, to extraction and analysis the high frequency component inthe “Interesting Area”, judge the defocus blur degree of image; thirdly calculating the averagegrey value in the above of the pupil to discard the seriously occluded images by eyelid; lastlythe occluded degree by eyelash is measured through the horizontal high frequency componentin the “InterestingArea”.In the iris image pre-processing link, in order to reduce the computational complexityand positioning time-consuming, this paper proposes a new localization algorithm based onthe gray vertical projection and the mathematical statistics: Firstly, using the least squaresmethod and the binary image to complete the pupil localization; secondly, using Cannyoperator for edge detection; then through gray vertical projection to screen the coordinaterange of center of the iris and eliminate the noise information; Finally by using themathematical statistics method to derive the circle Parameters, and complete the iris outerlocation; using Daugman’s Rubber Sheet Model and bilinear interpolation method to normalize the iris image to transform the annular iris image into the same size rectangle image;in order to cut the eyelid and eyelash interferences impact, we select the “Interesting Area” atthe bottom of pupil normalized iris as the basis of feature extraction and pattern matching.Compared with the Daugman and Wildes algorithm, the proposed algorithm stand out inspeed, accuracy and robustness.Because of the uniqueness and randomness of the iris image, feature extraction operationis important. This paper propose an effective method based on the statistical characteristicsCS-LBP operator, it improve the traditional CS-LBP operator to reduce the dimensiondescription code and the memory consumption, not only have the gray scale invariance, butalso can extract the iris texture information in different direction, combined with theHamming Distance match method, get a fast and effective method of feature extraction. Inaddition, the paper based on the False Accept Rate (FAR) and the False Rejection rate (FRR),compare and analyze the value of parameter N and R of CS-LBP operator, propose theremedial measures for the phenomenon of rotation angle. experiment results on CASIAdatabase demonstrate that this method is ideal．...