Automated Fingerprint Identification Technology Research

Along with development of society, traditional security systems which based ontoken or password cannot meet the demand of modern society. Fingerprint recognitiontechnique that is safer and has more feasibility was appeared under this situation.This article's background is the system of the family-door's lock which is basedon the technique of automatic fingerprint identification, so the algorithm must beefficient, taking small storage memory, and can be loaded into embedded systembased on ARM9. The work flow of Automatic Fingerprint Identification System canbe described as follows: the preprocessing of the original fingerprint image, extractingof the minutiae, and matching. This article emphasizes on the research of the contentbelow and proposes correlative improves method.a) Detection of singular points. On the research of the orientation featureadjacent to the singular points, this article advanced a fast method to find the probableareas where the singular points may exist in. Then use the Poincare Index algorithm tofind the exactly position of two kinds of singular points: core points and delta points.Then designed a template to detect the false singulars.b) Extraction the minutiae. We use the characteristic of the 8 adjacent pointsaround the minutiae to extract the two kinds of minutiae. To different kinds of falseminutiae, this article discussed different methods to detect them and adjusted themethods.c) Matching of Fingerprints. This thesis proposes an improved minutiaematching algorithm. In the first step, we used the concept of vector triangle to findsome reference points for the next step. Then rectify the angle of minutiae, match theminutiae from two fingerprints in polar coordinate, we use the double-boundary boxto determine whether two points is matched or not. According to each twocorresponding reference points, how many pairs of points can be matched, and makeit as the score of matching. Then get the maximum, and make the current referencepoints as the last result datum. On condition of a fingerprint don't have muchminutiae, this article integrated the algorithm based on ridge similarity.