摘要
大部分手指静脉识别算法使用单个特征识别,识别的基本是静脉区域特征。研究发现背景区域对静脉识别也有辅助作用,为此,提出一种融合静脉曲率灰度特征、曲率细线特征及背景曲率灰度特征的阈值识别算法。用高斯模板对图像进行曲率特征提取和分解,得到背景曲率灰度特征和静脉曲率灰度特征,再通过提取曲率特征得到曲率细线特征,并用阈值融合算法将3种静脉图像特征融合起来进行识别。实验表明,所提算法的性能优于基于方向滤波提取细线特征的算法和基于Hessian矩阵提取细线特征的算法。训练样本中,认假率为10-6时,其拒真率最低降至9.13%,性能显著提高。
Most of the finger vein recognition algorithms use single feature recognition,and mostly recognize the vein region feature.Study shows that the background area is also helpful to the vein recognition.Therefore a threshold recognition algorithm for fusing vein curvature grayscale（VCG）feature,curvature thin line（CTL）feature and background curvature grayscale（BCG）feature is proposed.Gauss template is used to extract and decompose the curvature feature of the image,achieving BCG feature and the VCG feature.Then CTL feature is obtained by extracting curvature feature,and three kinds of vein image features are identified by the threshold fusion algorithm.The experimental results show that the proposed algorithm has a better performance,compared with the algorithm based on the direction filtering extracting thin line feature and the algorithm based on the Hessian matrix extracting thin line feature.In the training sample,the false rejection rate（FRR）is reduced to 9.13% when false accept rate（FAR）is 10^（-6）,and the performance is improved significantly.
作者
蓝师伟
沈雷
LAN Shiwei;SHEN Lei(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China)
出处
《杭州电子科技大学学报（自然科学版）》
2018年第5期24-28,84,共6页
Journal of Hangzhou Dianzi University：Natural Sciences
基金
中国博士后科学基金资助项目(2014M562302)
浙江省新苗人才计划资助项目(2017R407077)
关键词
指静脉识别
曲率
细线特征
阈值融合算法
finger vein recognition
curvature
thin line feature
threshold fusion algorithm
作者简介
通信作者：沈雷，副教授，研究方向：大规模天线、认知无线电。Email：shenlei@hdu．edu．cn。
