北京航空航天大学学报 ›› 2015, Vol. 41 ›› Issue (1): 110-116.doi: 10.13700/j.bh.1001-5965.2014.0453

• 论文 • 上一篇    下一篇

基于ASIFT的离线签名认证方法

唐有宝1, 卜巍2, 张恩泽3, 邬向前1   

  1. 1. 哈尔滨工业大学 计算机科学与技术学院, 哈尔滨 150001;
    2. 哈尔滨工业大学 媒体技术与艺术系, 哈尔滨 150001;
    3. 哈尔滨工业大学 航天学院, 哈尔滨 150001
  • 收稿日期:2014-04-28 出版日期:2015-01-20 发布日期:2015-02-04
  • 通讯作者: 卜巍(1977-),男,黑龙江哈尔滨人,讲师,buwei@hit.edu.cn,主要研究方向为图像处理、模式识别、数字艺术设计. E-mail:buwei@hit.edu.cn
  • 作者简介:唐有宝(1987-),男,湖南衡阳人,博士生,tangyoubao@hit.edu.cn
  • 基金资助:

    国家自然科学基金资助项目(61073125,61350004);中央高校基本科研业务费专项资金资助项目(HIT.NSRIF.2013091,HIT.HSS.201407)

Offline signature verification based on ASIFT

TANG Youbao1, BU Wei2, ZHANG Enze3, WU Xiangqian1   

  1. 1. School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China;
    2. Department of New Media Technologies and Arts, Harbin Institute of Technology, Harbin 150001, China;
    3. School of Astronautics, Harbin Institute of Technology, Harbin 150001, China
  • Received:2014-04-28 Online:2015-01-20 Published:2015-02-04

摘要:

通过分析现有局部不变特征方法及含伪装签名数据集的特点,提出了一种基于具有仿射不变性的尺度不变特征变换(ASIFT)的离线签名认证方法.该方法对签名图像进行预处理(包括灰度化和放缩),对处理后的图像进行ASIFT关键点检测和特征描述符提取,对从查询签名图像和参考签名图像中提取到的描述符进行匹配,对匹配的结果采用随机采样一致性(RANSAC)方法去掉错误匹配,并计算正确匹配点的描述符之间的平均距离;通过比较平均距离及正确匹配点的个数与给定阈值的大小来判断认证是否成功.使用了含伪装签名的数据库对提出的方法进行测试,实验结果表明该方法与现有方法相比等误率降低了5%.

关键词: 离线签名认证, 伪装签名, 局部特征, 仿射且尺度不变特征变换, 随机采样一致性

Abstract:

A novel offline signature verification method based on the affine and scale invariant feature transform (ASIFT) was proposed by analyzing the existing local invariant features and the property of the disguised signature. The method consists of the following steps, the preprocessing including image graying and resizing was performed on the signature images; the key points were detected and the corresponding descriptors were extracted from the processed images; the descriptors extracted from the query and reference images were matched and the random sample consensus (RANSAC) algorithm was used to refine the matched result. Then the average distance was computed according to the distances between the descriptors of the correct matched points. The verification decision was given by comparing the average distance and the number of the correct matched points with the thresholds. The proposed method was evaluated on a public signature dataset including the disguised signatures and the experimental results show that the proposed method outperforms the state-of-the-art algorithms with reducing the equal error rate (EER) by 5%.

Key words: offline signature verification, disguised signature, local feature, affine and scale invariant feature transform (ASIFT), random sample consensus (RANSAC) algorithm

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发