The Research On Matching Algorithm Of Fingerprint Identification

Fingerprint identification technology，which is an integrated technology usingfingerprint for automatic recognition，is a research focus in the area of biometricrecognition. Application of fingerprint identification has a long history. A lot ofresearehers are working hard on fingerprint identification technology and propose anumber of effective algorithms with fruitful results. However，there are still someproblems unresolved. The expanding applications and the diversity of fingerprintidentification bring some problems such as high computation complexity， lowmatching speed and accuracy of fingerprint matching algorithms. In the thesis wepropose two fingerprint matching algorithms to improve the accuracy, complexity andcomputing speed.Firstly a fingerprint matching algorithm based on the Haar wavelet transform isproposed. This scheme directly extracts fingerprint features from the binaryfingerprint image and matches the features extracted previously by included anglecosine method. This method effectively simplifies some complicated pretreatmentsand greatly improves the speed of matching algorithm because it directly extractsfingerprint features from the binary fingerprint image. Experiment indicates that thisfingerprint matching algorithm has nice translation invariant and rotation invariant. Ina word, this fingerprint matching algorithm is feasible with characteristics of easyimplementation, good real-time and high robustness.To improve the accuracy of fingerprint matching algorithm and decrease the costtime， we propose a Segmented Point Matching Algorithm (SPMA) based oneigenvectors in this paper. This method is an improved variance of a SPMA proposedby Dinesh P Mital etal, which is a minutiae matching approach based on re-matchingmethod.In the proposed approach,we get the local matching scores infirst-local-matching step after basic minutiae extraction and purification task. Thenthe local matching scores are applied to adjust two images and do asecond-global-matching step. The final decision is made from the matching rate of thematching vectors in the second matching step. The approach is supported byexperiments on three test collections, including both international fingerprintverification competition databases and the database collected by on-line FPS200capture system. Experimental results show that our approach is accurate and effective performance.