Research Of Automatic Fingerprint Dentification System

Fingerprint is a kind of inherent biologic feature which have advantages of difficult loss and avoiding fake. It can provide users a safe and reliable attestation approach and applies in many fields. The automatic fingerprint recognition performance can not satisfy users because of the disadvantages of noise in fingerprint image and skin flexibility. Therefore, the research of automatic fingerprint recognition system is important.The components of automatic fingerprint recognition system are investigated in this paper which mainly include preprocessing, fingerprint feature extraction and fingerprint feature matching.The fingerprint preprocessing section is investigated in this paper. Five proportions of fingerprint preprocessing are discussed which include fingerprint normalization, fingerprint image segmentation, fingerprint imge enhancement, fingerprint image binarization and fingerprint image thinning. The segmentation algorithm of fingerprint image, Gabor filter enhancement algorithm and OPTA thinning algorithm are investigated emphatically. The preprocessing flow is improved and quality evaluation of fingerprint image is introduced, which improve preprocessing efficiency.The fingerprint feature extraction section is investigated in this paper. Fingerprint feature is classifed and summarized, and the fingerprint feature extraction algorithms are choosed according to local and global feature. Finally, the spurious feature point elimination algorithm is investigated and a new method based on framework density and contrast difference is proposed for blurry fingerprint image.The fingerprint feature matching section is investigated in this paper. Minutia point registration algorithm is firstly discussed. The fingerprint feature matching algorithm is investigated emphatically, and the disadvantage of miutia matching method based on point mode is analyzed. The matching algorithm based on triangle fiducial point location is choosed to solve above problem. This algorithm overcomes the shortcoming of dependance to center, introduces a defination of boundary box in matching course and improve the condition of multi-judge, which improve matching rate.