一种基于信息融合的指纹奇异点提取及纹型分类算法  

A FINGERPRINT SINGULAR POINTS EXTRACTION AND CLASSIFICATION ALGORITHM BASED ON INFORMATION FUSION

作　　者：张晔[1] 彭小奇[1,2] 钟云飞[3] 唐英[1] 

机构地区：[1]中南大学物理与电子学院,湖南长沙410083 [2]湖南第一师范学院信息科学与工程系,湖南长沙410205 [3]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《计算机应用与软件》2016年第4期246-251,共6页Computer Applications and Software

基　　金：湖南省自然科学基金重点项目(10JJ2048);湖南省科技厅项目(2011GK3079)

摘　　要：准确可靠地实现纹型分类对提高大容量指纹库中的检索和匹配效率具有重要意义。提出一种基于信息融合的指纹奇异点提取与纹型分类算法。首先,分别给出一种基于奇异点区域方向场信息和奇异区复数滤波场信息的改进的奇异点提取算法,并将两者融合以完整提取奇异点;再利用所提取奇异点邻域的Gaussian-Hermite矩的分布属性剔除伪奇异点;最后,利用奇异点的数目和位置关系及中心点的主方向将指纹分为常见的六种纹型,对缺少三角点的指纹,使用脊线跟踪算法进行分类。实验表明,该方法新颖有效,具有较高的准确性和鲁棒性。Accurately and reliably classifying the fingerprint images into different classes is very important to promote the retrieval and matching efficiency in large fingerprint databases. In this paper we propose an information fusion-based fingerprint singular points extraction and classification algorithm. First,we present two improved singular points extraction algorithms,which are based on the orientation field information of singular points region and on the complex filtering field information of singular regions respectively,,and fuse these two algorithms to extract the singular points in whole; Secondly,we employ the distribution attribute of Gaussian-Hermite moments in neighbourhood of the extracted singular points to cull the false singular points; Finally,we classify the fingerprints into six familiar classes by making use of the number and the location relation of singular points as well as the main direction of core singular points,. For the fingerprints lacking the delta singular points,the ridge tracing algorithm is adopted. Experiment shows that the proposed fingerprint singular points extraction and classification algorithm is novel and effective with higher accuracy and robustness.

关 键 词：指纹 奇异点 信息融合 主方向 分类 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]