基于特征融合的指纹质量评估算法  

Evaluation Algorithm of Fingerprint Quality Based on Feature Fusion

作　　者：马丽红[1] 余德聪[1] 卢汉清[2] 陈志清[3] 

机构地区：[1]华南理工大学电子与信息学院,广东广州510640 [2]中国科学院自动化研究所模式识别国家重点实验室,北京100080 [3]广东省公安厅刑事技术中心,广东广州510050

出　　处：《华南理工大学学报（自然科学版）》2007年第5期20-24,共5页Journal of South China University of Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60472063);广东省自然科学基金项目资助(04020074/CN200402)

摘　　要：针对公安刑侦指纹库中指纹质量参差不齐的问题,提出一种基于指纹图的空间和频率特征融合的指纹质量评估算法.首先用中心区域黑像素比例、环形谱结构和Gabor方向特征来单独评估指纹质量,分析各种特征的评估性能.然后,根据各评估特征对不同质量类型指纹的评估特性,给出一种新的非线性指纹评估准则,它通过对3种质量特征的评估结果进行非线性融合加权来计算指纹的质量评估分数.在公安刑侦指纹库上测试了该评估准则,结果表明,文中提出的非线性融合评估方法能够准确地评估指纹质量.In order to solve the problem of quality difference in fingerprint in public security fingerprint database, a new fusion method to evaluate fingerprint quality is proposed by combining the spatial and the frequency features of a fingerprint image. In this method, the black-pixel ratio in the central area, the ring structure of Fourier spectrum and the directional Gabor features are first employed to respectively assess the fingerprint quality, and their corresponding evaluation performances are discussed. Then, according to the evaluation capacities of these features for different quality images, a nonlinear fusion approach is presented to obtain the final quality scores by weighing the nonlinear fusion of the features mentioned above. The fusion criterion is finally tested in a public security fingerprint database, the results demonstrate that the proposed nonlinear fusion algorithm is feasible in fingerprint quality evaluation.

关 键 词：指纹 质量评估 算法 非线性融合 GABOR特征 黑像素比例 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]