Research On Fingerprint Recognition Systems Under The Visual C++

The twenty-first century, biotechnology and information technology are developing rapidly and becoming more mature, by means of biometric identity recognition technology has also been a rapid development. In particular, modern fingerprint identification technology has become as a more reliable and widely used biometric technology, the fingerprint is unique, and stability, players of the advantages, which makes fingerprint identification technology is superior to other biometric identification technology.At present, although the market has been forming a number of fingerprint identification products, but the fingerprint recognition algorithm needs to be improved, affected by realistic environment, as well as many commercial interests involved no public literature, the ideal fingerprint recognition technology is still a very difficult Research.In this paper, the author does some researches for the fingerprint recognition of the various parts, describes existing algorithms, and makes some improvements, which include fingerprint preprocessing, feature extraction and fingerprint matching parts. In the fingerprint preprocessing part, fingerprint image normalized, image segmentation, fingerprint image enhancement, image binarization and thinning algorithm are studied and comparing various algorithms, the algorithms are selected for better fingerprint images Pretreatment.In the feature extraction part of this thesis, the eight neighborhood encoding algorithm is used to extract singularities. As a result of extracting feature points of the algorithm there is a lot of pseudo-feature points, which will affect the subsequent fingerprint matching. Therefore, we use the pseudo ridge tracking feature points and filter out the pseudo-feature. Finally, save the fingerprint feature point information, laid the foundation for the fingerprint matching.For matching, the most commonly used method is to match the details presented by the FBI point model, the matching fingerprint identification into a point pattern matching. This paper also uses this matching method, and calibration of ridge minutiae matching algorithm has been improved, that is, match point in the details, taking into account more details of the ridge line of the message near the point. The traditional feature matching algorithm is to compare the distance of two points to their reference point whether is within the scope of the default to determine whether the two feature points match, there is no effective use of ridge information. This approach is the use of a ridge line full of information, the first point on the ridge details of sampling, then the similarity comparison, the results as an important reference point for the selected criteria. This improves the accuracy of the selected reference point, reducing the number of false matches to improve the matching accuracy and speed.