High Resolution Fingerprint Pore Extraction And Matching Algorithm Research

With the development of auto fingerprint identification system application,advanced requirement is presented concerning system security. Bottlenecks are alreadyappearing in traditional fingerprint identification system. Meanwhile, the performanceof fingerprint extracting device has been improved significantly due to the rapiddevelopment of science and technology. This makes the third-layer fingerprint characterinformation such as pore possible. Consequently, the high resolution of fingerprintidentification becomes the focus.Compared with traditional fingerprint identification system, the high resolutionfingerprint identification system has evident advantages in terms of identificationaccuracy, prevent fraud, and bio-assay. However, the high resolution fingerprintidentification system also brings a lot of problems such as timeliness of fingerprintidentification. At present, most pore extracting methods are based on a single scale andsame direction models, which produce low performance while extractingmulti-dimension and directional pore. Therefore, improving the pore informationextraction accuracy is a key problem to the high resolution fingerprint identificationsystem. There are two pore matching algorithms. The first algorithm is based on thesecond layer finger character align and the second method is based on direct porematching algorithms. The first algorithm can get higher timeliness and the secondalgorithm can get higher accuracy. Solving the problem of timeliness and accuracy isfacing by the high resolution fingerprint identification system.An algorithm for pore extracting was proposed based on Gabor filter. Thisalgorithm made good use of multiple scale and direction characteristics of Gabor filter.The scale and direction factors were added into the Gabor filter to generate the Gaborpore model. The algorithm solved the single scale and isotropy problems of poreextraction, and the pore extraction accuracy was higher compared with other methods.A pore matching algorithm was also proposed based on minutiae pore selection.This method used minutiae pores as a selection policy and it choose some pores to porematching. This method was more accurate when compared with methods based on porealign and decreased the time complexity.Validation experiments were performed based on the high resolution fingerprint database that downloaded from the Hong Kong polytechnic university. The resultsshowed that the pore extracting and pore matching algorithms proposed in this thesisimproved the performance of pore extraction algorithms that currently used.