Research And Implementation Of Automatic Fingerprint Identification System Algorithm

Fingerprint identification has become one of the most popular, most convenient and most reliable person identification technologies since fingerprints are unique and inflexible as well as fingerprint identification technology is feasible and practicable. Fingerprint identification technology has report remarkable achievements and been applied to a wider field in the study during the last two decades. Recently, commercial products for fingerprint identification have been developed but could not meet some specific application performance requirements. Meanwhile, most research results are not made public to protect business interests. Therefore, this thesis presents an in-depth study on the key algorithm of fingerprint identifycation technology on the basis of an investigation upon the relevant documents of fingerprint identification systems.In the aspect of the preprocessing, a preparatory splitting of the original fingerprint image was carried out first and then normalized. Considering the directivity of fingerprint, the fingerprint image was enhanced throng directional image filtering and the enhanced fingerprint image was then binarized through partial dynamic threshold method. In the aspect of thinning, eight canceling templates were added to the improved OPTA thinning algorithm and new thinning algorithm was applied to achieve promising results.The extracted feature points include a large amount of pseudo feature points. Therefore, a detailed discussion about the pseudo feature points filtering algorithm was presented based on ridge tracing, which filtered most of the pseudo feature points and improved the discrimination ratio.For the matching, two algorithms were realized and compared. The experiment results were studied and the matching method based on triangle reference point locating was introduced. Effective improvement was made upon the original algorithm and the robustness of the algorithm was enhanced and the discrimination ratio was improved. In this thesis, the whole algorithm of automatic fingerprint identification system is realized on a personal computer with VC++ 6.0. With the purpose to prove the practicality of the algorithm, many tests were carried out on the platform built for the fingerprint identification software. Test results showed that the fingerprint identification algorithms applied or improved realized the function-ns well and met the basic practical requirements.