Combining singular points and orientation field for fingerprint classification
Chengpu Yu,Mei Xie,Jin Qi
2008-01-01
Journal of Computational Information Systems
Abstract:In this paper, we presented an algorithm to classify fingerprints pattern based on combination of singular points (core and delta points) and orientation field. Singular points of fingerprints can be extracted by Poincare index method, however arch fingerprint either has no singular points or has a core point and a delta point closely located at the center of the fingerprint, thus it can be identified by its particular condition. As singular points in many fingerprints are not completely collected for left loop, right loop and whorl pattern, so that we can use angular histogram of fingerprint orientation field in a special rectangle below core point as the classification feature to solve this problem. After getting angular histogram, we can classify these fingerprints through combination of BP network and SVM classifiers, which can not only supplement their advantages, but also improve their computation efficiency. At last, we experiment this algorithm on database of FVC2004, and the experimental results show that our classification accuracy is 93.2%, which is comparable with that achieved by currently prominent fingerprint classification methods. © 2008 Binary Information Press.