| Automatic fingerprint identification technique has become a research focus in the area of science in last decades. Much work has been done on fingerprint identification, but some technical difficulties still baffle the development of fingerprint identification, especially non-ideal fingerprint identification. Fingerprint matching and feature extraction are two main steps in fingerprint identification, in the process of fingerprint matching, under the theory of machiner learning, using the information got during the matching process, we improve the information about the template, and plus the type of information used in fingerprint matching to add the information which can be used in fingerprint identification. While in the feature extraction process, we improved the quantity of fingerprint orientation to reduce the noise exist in the information after feature extraction.Our work is focus on the template improvement, combined fingerprint matcher and orientation calculation.For fingerprint template improvement, based on the case-based method of machine learning and probability theory, we improve the template by deleting pseudo minutia, recovering losing genuine minutia and updating the information of minutia such as location and direction. Special fingerprint image database is built for this work. Experimental results on the new database and fvc2000 indicate that our method is effective and the quality of fingerprint template is improved evidently. Accordingly, performance of fingerprint matching is also improved stably along with the increase of using time.For combining matcher: the limitation of information used in fingerprint matching is the one of the most important problems during fingerprint matching. In this paper, a new kind of fingerprint matching based on the global orientation information is studied. And a combined fingerprint matcher based on the global orientations and minutia is also introduced. The result reported in the bac2005 indicates the valid of proposed matching... |
PDF Full Text Request