一种基于方向链的指纹分类算法  

A fingerprints classification algorithm based on orientation chains

作　　者：魏鸿磊[1] 刘丹妮[2] 

机构地区：[1]大连水产学院机械工程学院,辽宁大连116023 [2]东北大学东软信息学院计算机科学系,辽宁大连116024

出　　处：《光电子．激光》2008年第10期1410-1414,共5页Journal of Optoelectronics·Laser

基　　金：辽宁省博士启动基金资助项目(20071066)

摘　　要：提出一种基于方向场分析的指纹分类方法。以poincare方法定位指纹中心点,以其为起点,在两个相反的方向上对指纹的纹理方向进行追踪,形成两条方向矢量链,简称为方向链。不同类别指纹具有不同形状的方向链,以方向链总角度差和分支角度差作为分类依据,实现指纹的分类。该方法对指纹的平移和旋转具有无关性,计算量小,易于实现。在NIST-4数据库的50幅图像上测试取得了96%的正确率,在FVC2000的DB1和DB2、FVC2002的DB1和DB2 4个库的800×4幅图像上测试分别取得了89.1%、92.8%、92.3%和89.8%的正确率。A fingerprint classification algorithm based on orientation chains is presented. The core point is extracted in the orientation field using the poincare method firstly,then using the core point as start point,the orientation vector chains which is simply named as orientation chain is extracted by chasing the texture trend from two opposite directions in orientation field. The algorithm performs the classification based on the angle differences in whole chain and two branch chains. The classifier is invariant to rotation and translation and it is simple with less computation and easy to realize. The classifier is tested with 50 images in the NIST-4 databases and on 800 × 4 images in the four databases of FVC. For the NIST-4 databases,classification accuracies of 96% for the four-class problem are achieved,and for the FVC2000_DB1 database,FVC2000_DB2 data base,FVC2002_DB1 database and FVC2002_DB2 database, classification accuracies of 89. 1%,92. 8%,92. 3%, 89.8% for the four-class problem were achieved, respectively.

关 键 词：指纹识别 指纹分类 方向链 方向场 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]