基于曲率尺度空间的指纹特征提取算法  

Fin9erprint Feature Extraction Based on Curvature Scale Space

作　　者：陈书贞[1] 张俊军[1] 

机构地区：[1]燕山大学信息科学与工程学院

出　　处：《电子技术（上海）》2008年第1期99-103,共5页Electronic Technology

摘　　要：提出了一种基于曲率尺度空间(Curvature Scale Space,CSS)的指纹特征提取方法,它不同于经典的基于细化的算法。首先使用Canny算子得到增强后指纹脊线的轮廓,接着计算轮廓中每点的曲率,找出局部曲率最大点作为候选的细节点,然后使用自适应阈值和支撑域(Region of Support,ROS)删除由噪声产生的虚假点;最后利用指纹的细节点附近的轮廓和灰度信息,确定细节点的方向和类型,并使用得到的方向和类型进一步删除虚假细节点。该算法在FVC2002指纹数据库中进行测试,实验结果表明,该算法计算复杂度低,得到的指纹细节点精确、稳定,因此适合实际应用。A feature extraction method using the curvature scale space for fingerprint image is proposed.,which is distinct from thinning based approach.Firstly,Canny operator is used to extract the contour of ridge,then curvature value is calculated at every point in the contour, subsequently,local maximums are found as the candidates of minutiae.Secondly,some false minutiae that are caused by noise can be removed by adaptive threshold and region of support.Lastly, the orientation and style of minutiae are decided by the contour and gray information at the neighborhood of minutiae and lots of rudimental false minutiae can be removed by the information of orientation and style.This algorithm was tested on the fingerprint database of FVC2002,the result show that this approach is efficient in computation,minutiae obtained is accurate and stable,therefore suitable for practical applications.

关 键 词：指纹细节点 曲率尺度空间 GABOR滤波器 指纹增强 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]