Research On Algorithms Of Iris Recognition And System Implementation

Information security and identity recognition are becoming universal problemand getting more and more attentions with the fast development of Internet technology.Traditional identity recognition methods cannot satisfy the needs in accuracy,securityand practicability.Biometric technologies using human physical characteristics havesome advantages in robustness,convenience and anti-falsification and give us analternative.Among numerous biometric technologies,iris recognition has the bestperformance and is getting more attentions.This thesis first gives detailed analysis and conclusion to each step of an irisrecognition technology,then studies some crucial technologies.The thesis alsoproposes some new ideas and algorithms based on above work and designs aneffective and practical iris capture and recognition prototype system.The maincontributions and novelty of our work are as follows:1.Proposed an iris location method based on multi-threshold and local maximumradial gradient.This method adopts different strategy to locate inner and outerboundaries separately.Since the intensity distribution in pupil area is not uniformcommonly,and it is difficult to separate the pupil accurately with one universalthreshold.So for inner boundary (i.e.pupil boundary)location we use amulti-threshold based method which changes threshold every time from minimumvalue to segment pupil until the whole pupil area is found.This strategy not only canget more integrated pupil region but also exclude background as much as possibleIn the process of iris outer boundary location,the image is translated intorectangular region taking pupil center as origin at first.Then calculate the verticalgradient of each column and find the maximum.From one maximum point we canfind the local maximum gradient of next column and link those points to form somelists based on location rules.Finally the outer boundary parameters are got byselecting the longest list as iris outer boundary.The method is simple,accurate,robustand efficient since its searching space is N~2 related.Experiment results show thatthe speed of the proposed method is 500 times as fast as integral differential operatorused in CASIA vl.0.2.Proposes an algorithm based on mask templates for eyelashes and eyelidsdetection.This method designs eight different mask templates according to theocclusion degree of eyelids and eyelashes.In order to estimate the occlusion degree,two areas with 90 degrees span separately where eyelashes and eyelids often appear in the normalized iris image are selected and divided into eight blocks with different size.Then select two areas with less eyelashes and eyelids in left and right parts of thenormalized iris image as references to calculate their mean intensity and variance.Then calculates the intensity second moment of each block relative to reference areasand determine whether the block is occluded by eyelids and eyelashes.Finally selectsan appropriate mask template according to the position of the block which is occludedby eyelids and eyelashes.The method is simple,robust and can determine occlusiondegree efficiently and improve the performance of iris recognition system.3 Proposes a segmentation based method for iris feature extraction using pulsecoupled neural network (PCNN).This algorithm uses local irregular blocks asfeatures directly and uses PCNN to segment these blocks out.Finally estimates thesimilarity of two iris codes by calculate their Hamming distance.This method is fast,efficient,robust and easy to implement and test.4.Designing a practical iris capture and recognition prototype and discussingabout the difficult points and resolutions carefully based on above works.Then givethe detailed description about camera selection,lens requirement,auto focusing anddividing strategy for software/hardware in an iris capture/recognition system.Andbriefly discusses about the hardware designing proposal.5.Introduces the software design proposal of iris capture/recognition system anddiscusses about the firmware programming and USB driver program designing formotor driving circuit.And give the detailed designing proposal of iris recognitionsystem.