Research On Fingerprint Recognition Algorithm

Fingerprint recognition is a method to identify ID by inherent biometric characteristics. Comparing to others methods, automatic fingerprint recognition is a technique with better ID confirmation, not only having many advantages believed nasty and safe, but being carried out easily and meaningfully. Currently, the automatic fingerprint identification system (AFIS) has already carried out in numerous fields, such as ID confirmation, E-business, Internet safety, gate control, information safety, distant attestation etc. Although having a extensive application, fingerprint recognition still has many problems to be resolved, such as the applied catholicity, accuracy, efficiency of the algorithm and credibility, etc. The fingerprint recognition still would conduct and actions a long-term study topic existence. This kind of research has higher meaning and value.After investing and analyzing the existent fingerprint recognitions and algorithms, this thesis emphasizes studying some main algorithms in AFIS, such as the direction image, enhancing, thinning, features'extraction, fingerprints'matching and carrying out a set of integrity of the fingerprint recognition algorithms. To obtain orientation image, this thesis proposes a fast algorithm based on gray difference, which utilize direction template to acquire orientation directly. For enhancing, a novelty filtering method is proposed, which integrates median filtering and trend-pass filtering, having a good effect in binarization. In addition, for improving Gabor filtering, this thesis proposes a self-adapting Gabor filtering algorithm which reduces enhancing time greatly. About the binarization in fingerprint, this thesis proposes a modified morphologic thinning algorithm and obtains skeleton drawing with single pixel. For eliminating the false feature, this thesis proposes two methods, one is based on orientation and frequency, and the other syncretizes little frame and tracing ridges, which improves the speed of operation and effect in elimination. Finally, in fingerprint matching, this thesis utilizes structure characteristic information by characteristic point and its surroundings'to build up the characteristic matrix to carry on matching an operation, proposes a matching algorithm based on structure characteristic information, which has a good robustness to circumrotate, excursion, distort and compress..All the algorithms mentioned above, this thesis has programs and realizes them with MATLAB7.0 on PC. The result of emulation experiment indicates that adopting these algorithms can function well in every work and come to the basic request of fingerprint recognition. These algorithms in this thesis provide a good theories basis and the experiment foundation for further research of developing AFIS.