Research Of Fingerprint Preprocessing And Recognition Algorithm

At present, with the development of e-commerce, the traditional password or token based security systems can not meet the social needs anymore. Thus, a more secure system that is more powerful in anti-counterfeiting is needed. Biometrics, due to its uniqueness and permanence, is very suitable as the basis for personal identification.Among the numerous biometrics, fingerprint recognition technology has become a hot research field in pattern recognition due to its unique advantages, and it has exhibited bright market prospects. For constructing a fingerprint recognition system, fingerprint image preprocessing is the first step. The result of the preprocessing step has direct impact on the performance of entire system. In this paper, the fingerprint image preprocessing technology is studied, including segmentation, enhancement and thinning. First, the collected fingerprint images were segmented by calculating intensity field in the segmentation step. And then for the enhancement and thinning part, the author implemented a variety of algorithms and analyzed their advantages and shortcomings based on the study and induction of other scholars'achievements both at home and abroad. Besides, some improvements were made on part of those algorithm, the improved algorithms are enhanced both in efficiency and processing results. The improved algorithms are as follows:(1) For fingerprint image enhancement algorithms, two kinds of existing algorithms were studied: spatial domain enhancement and frequency domain enhancement. Both the median filtering and STFT based enhancement algorithms were implemented. The Gabor filter based fingerprint image enhancement algorithms were analyzed and some improvements were made, a Gabor function enhancement algorithm based on wavelet domain was proposed. By decomposing the 2-D Gabor function into two 1-D functions, the computational complexity was reduced and the efficiency was improved.(2) For the aspect of fingerprint thinning, the common used OPTA thinning algorithms (both the typical one and the improved one) and fast thinning algorithm were introduced and implemented. Aims at the problems that the thinning result at bifurcation points can not be thoroughly and glitches will be produced by using the OPTA algorithms, and that the result of fast thinning algorithm can neither be thoroughly, a new direction image based thinning algorithm is proposed. Experimental comparison showed that the proposed algorithm is better than the previous two algorithms.Finally, in order to show the impact of preprocessing step on the entire fingerprint recognition system, some typical feature points extraction algorithms and the triangle topology based matching algorithm were introduced.In this paper, the entire fingerprint recognition system was designed using Visual C++ 6.0, and the various algorithms introduced above have been implemented on this platform, the proposed algorithms proved to be efficient, this has laid a solid foundation for further study.