基于伪Zernike矩和Hough变换的脱机中文签名鉴别  

Off-line Chinese Signature Verification Based on Pseudo-Zernike Moment and Hough Transform

作　　者：陈万军[1] 鲁继文[1] 梁敏[2] 

机构地区：[1]西安理工大学,陕西西安710048 [2]山西财经大学信息管理学院,山西太原030006

出　　处：《现代电子技术》2007年第12期105-107,114,共4页Modern Electronics Technique

摘　　要：利用伪Zernike矩和Hough变换提取了脱机中文签名图像的静态特征和动态特征,采用加权欧氏距离分类器完成签名鉴别。在690个真伪签名的较大规模样本库上进行测试,系统最高正确识别率为87.0%。利用签名图像不同特征能提供信息互补的特点,在决策层上进行了特征融合识别。系统在保持对伪样本拒绝率为71%的情况下,对真实签名的正确识别率仍可达80.4%。实验结果表明,多特征信息融合方法能较好地提高签名鉴别系统的识别性能。Pseudo- Zernike moment and Hough transform are adopted to extract the static features and dynamic features for off- line Chinese signature image respectively,and signature verification system has been completed based on the weighted Euclidian distance classifier. This method has been tested on a database consisting of 690 signature image samples and the highest correct recognition rate of this system is 87.0%. Then,feature information fusion recognition on the decision level has been performed by the fact that different features can provide mutual- complement information, whose correct recognition rate still achieves 80. 4% while the forgeries rejection rate is 71%. Experimental results show that multiple feature information fusion can improve the performance of signature verification system effectively.

关 键 词：脱机签名鉴别 伪ZERNIKE矩 HOUGH变换 信息融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]