The Fingerprint Preprocessing And Matching Algorithms

In recent years， the biometric recognition technology has been widely applied inmany fields，such as the access control system with iris recognition，fingerprint punchcard machine， digital signatures， etc. The AFIS (Automatic Fingerprint IdentificationSystem) which is an important branch of biometric recognition， has been widelyapplied and studied due to its permanence， reliability， maturity and low-cost.After the study of fingerprint identification system for nearly four decades，a greatdeal of progress has been made， however， there are still bottlenecks in the algorithm，such as the performance of the algorithm is not so good enough that the hardware doesnot achieve the real-time performance; low quality fingerprint image recognition cannot recognized so as to extract the minutiae effectively. On the basis of previousstudies， we effectively improve the performance of the automated fingerprintidentification system by the improvements of the preprocessing algorithm and theoptimization of the matching algorithm.The main contents are listed as follows:(1) Study the fingerprint segmentation algorithm. Firstly， we proposed the basicfingerprint segmentation algorithm based on statistical features， and it is by calculatingthe grayscale variance, mean and gradient of the fingerprint for image segmentation.Secondly, a fingerprint segmentation method based on morphology has adopted. Themethod can effectively remove the noise in the image，broken ridges by open operationand closing operation(2) Analyzing the traditional fingerprint enhancement algorithm----the imageenhancement algorithms based on Gabor filter. This algorithm on the direction of thefield filter is relatively time-consuming and can produce false ridge. Therefor， weproposed the ant colony enhancement algorithms on the basis of bionics， the algorithmwhich simulated the proceeding of the ants’ finding food can effectively improve thequality of the fingerprint image.(3) In order to extract effective feature points in the fingerprint image, weintroduced a synthetic thinning algorithm based on the combination of the advantages ofthe fast thinning algorithm and the improved OPTA thinning algorithm. (4)A matching algorithm based on the principle of similar triangles is introduced. Itdetermined whether the two fingerprint images are matched or not by comparing thenumber of similar triangles，and greatly improved the accuracy of the match.