Genetic Algorithm Applied To Slight Distortion Fingerprint Matching

Fingerprint identification is the pioneer application of pattern identification as well as the most successful application of biology identification technology. Fingerprint identification which is provided with unique and reliable characters, synthetically exert sensor, biological, electronic, digital image-processing technologies. By now, fingerprint has been applied in more and more field concerned with individual identity distinguishing in daily life. The research and development of fingerprint identification system has become the hotpot not only in academe but also in business. Compared with other biological characters distinguishing technology, such as voice identification or iris identification, fingerprint identification possesses many original merits. For its practicability and feasibility, fingerprint identification has considered as the mainstream of biological characters identification in the future..Fingerprint identification system consists of four approaches: 1) fingerprint image acquisition;2) pre-processing;3) minutiae extraction;4) minutiae matching. Minutiae matching is the core step, for the performance of the match algorithm affect the identification capability, speed and efficiency directly. Many factors, such as shape distortion, dry or wet images, deformity, will depress the matching veracity. Then how to deal with the non-linearity image distortion in matching is a challenge in fingerprint identification. It's very important to improve the performance of fingerprint identification in both theory and application.The dissertation researches the development status and key problems of identification in detail, concludes existed matching algorithms; then discusses deeply on genetic algorithm; finally designs a re-matching method aiming at slight distortion fingerprint matching, combined genetic algorithm and delimitation box method. The scheme adopting genetic algorithm, minutiae matching theory and size-alterable delimitation box method, exercises two rounds matching processes. Compared with other re-matching methods' results, this algorithm has higher correct rate and lower false accepted rate. This fingerprint identification algorithm is based on minutiae matching, for minutiae is took for the unique and most believable character. This matching algorithm assumes that translate, circumrotate and size zoom exist between the input fingerprint image and the template fingerprint image. The matching algorithm adapts bogus minutiae creation and real minutiae lost in some range, also tolerates minutiae excursion in a small delimitation box.Finally, some pieces of advices on how to improve the performance of the matching algorithm for distortion fingerprint are given.