Based On Point Pattern Matching, Fingerprint Recognition Technology

This thesis's background is the system of the family-door's lock which is based on the technique of automatic fingerprint identification. The work of the system's software section achieves includes registering and identifying. The flow has such step as the preprocessing of the original fingerprint image, extracting of the minutiae, and matching. This thesis emphasizes on the research of the content below and proposes correlative improved method.a) Cutting of the original fingerprint image. Firstly, using different windows to smooth the pixel directional image we will obtain different result. Based on the Poincare index algorithm, find the position of the two kinds of singularities: core point and delta point. If there are some core points in the image, make their center as the image's centroid point. Otherwise, if there are some delta points, make their center as the image's centroid point. In the end, center on the centroid point, and cut the image to get a smaller image.b) Extraction the minutiae. This thesis proposes nine templates to determine two kinds of minutiae. Based on six kind of structure of the false minutiae, this thesis also proposes method to eliminate. Also we discuss the method to verdict the validity of the bifurcate and endpoint.c) This thesis proposes an improved minutiae matching algorithm. In the first step, calculate every two points' matching score to get a matrix, and then find some reference points for the next selection. The matching in polar coordinate, propose a new method to correct angle. According to every two corresponding reference points, obtain how many pairs of points can be matched, and make it as the score of matching. Then get the maximum, and make the current reference points as the last result datum. This thesis discusses the method to avoid unusual error in finding reference points. Next, compare the two lines which the reference points at. At last, use K-Nearest Neighbor algorithm to determine the matching result.d) Registering the fingerprint template. This thesis discusses two methods. One is to save three templates for a finger when it is pressed three times. The other is to save a compositive template for it. By analyzing result of experiment, we select the former.