基于邻域特征的笔迹鉴定算法的研究
|
摘要
作为计算机学科的一个重要应用方向,生物特征识别技术正在成为一个蓬勃发展的研究领域,笔迹鉴定技术则是其中具有吸引力的一个方向。近年来,一些发达国家已把生物特征识别技术,从研究阶段转向应用阶段,我国一些研究机构和企业也相继开展了这方面的研究开发工作,而且竞争十分激烈。另外,高准确率的签名笔迹鉴定是对个人、集体及国家利益不受侵犯的有利保证。个人签名被他人伪造后,有可能引起严重的社会后果或巨大的经济损失。对于我们这样一个人口大国,由于签名引起的纠纷案件越来越多,准确鉴定签名真伪的技术也显得越来越重要。因此,签名鉴定有着极其广泛的应用前景和重要的战略意义。
鉴于笔迹鉴定技术的研究现状,对其进行邻域特征分析和研究是可行的。本文主要完成了以下工作:
1)分析了已有的细化算法:快速并行算法和形态学细化算法,根据像素的邻域信息提出高精度并行手写体汉字细化算法;
2)在对已有的确定笔迹特征点算法研究的基础上,针对其不足,提出基于邻域特征的笔划交叉点提取算法;
3)结合人工神经网络和遗传算法的特点,提出基于遗传神经网络笔迹鉴定算法,实验证明了该算法的可行性和有效性。
最后阐述了作者在本课题研究过程中的心得和体会,指出了今后进一步探索和研究的方向。
As a major application direction, biometrics technology is becoming a fast growing research field, and the signature verification technology is one of the most attractive directions of it. Recent years, some developed countries have turned biometric identification technology from the research phase to the application stage. Some of the research institutions and enterprises in China have also began the work in this field, and the competition is very fierce. Besides, signature verification with high accuracy is a good guarantee to protect the interests of the individual, collectivity and nations from being infringed. Personal forged signatures may lead to serious social consequences or huge economic losses. For our country that has such large population, with disputes caused by signature arising, it is becoming more and more important to correctly identify the authenticity of the signature technology. Therefore, the verification of signatures has an extremely broad range of applications and an important strategic significance.
Based on the current study status quo of the signature verification technology, it is feasible to analysis and research on the features of its adjacency. The paper does several works as follows:
1) Analyze the existing thinning algorithm: fast parallel algorithm and mathematical morphology thinning algorithm, and propose the parallel thinning algorithm for handwritten characters with relatively high accuracy based on adjacency information.
2) On the basis of the existing algorithms, which determine the feature points of handwritten characters, and considering the disadvantages of them, this paper conducts studies on adjacency-based algorithm of extracting stroke joint points;
3) Combining the characteristics of genetic neural network and genetic algorithm, a handwritten characters verification algorithm based on genetic neural network is proposed in this paper, the feasibility and effectiveness of which have been proved by experiments.
Finally, the paper sets forth the author's self-experience in research into signature verification, and it also makes an outlook for further research direction.
鉴于笔迹鉴定技术的研究现状,对其进行邻域特征分析和研究是可行的。本文主要完成了以下工作:
1)分析了已有的细化算法:快速并行算法和形态学细化算法,根据像素的邻域信息提出高精度并行手写体汉字细化算法;
2)在对已有的确定笔迹特征点算法研究的基础上,针对其不足,提出基于邻域特征的笔划交叉点提取算法;
3)结合人工神经网络和遗传算法的特点,提出基于遗传神经网络笔迹鉴定算法,实验证明了该算法的可行性和有效性。
最后阐述了作者在本课题研究过程中的心得和体会,指出了今后进一步探索和研究的方向。
As a major application direction, biometrics technology is becoming a fast growing research field, and the signature verification technology is one of the most attractive directions of it. Recent years, some developed countries have turned biometric identification technology from the research phase to the application stage. Some of the research institutions and enterprises in China have also began the work in this field, and the competition is very fierce. Besides, signature verification with high accuracy is a good guarantee to protect the interests of the individual, collectivity and nations from being infringed. Personal forged signatures may lead to serious social consequences or huge economic losses. For our country that has such large population, with disputes caused by signature arising, it is becoming more and more important to correctly identify the authenticity of the signature technology. Therefore, the verification of signatures has an extremely broad range of applications and an important strategic significance.
Based on the current study status quo of the signature verification technology, it is feasible to analysis and research on the features of its adjacency. The paper does several works as follows:
1) Analyze the existing thinning algorithm: fast parallel algorithm and mathematical morphology thinning algorithm, and propose the parallel thinning algorithm for handwritten characters with relatively high accuracy based on adjacency information.
2) On the basis of the existing algorithms, which determine the feature points of handwritten characters, and considering the disadvantages of them, this paper conducts studies on adjacency-based algorithm of extracting stroke joint points;
3) Combining the characteristics of genetic neural network and genetic algorithm, a handwritten characters verification algorithm based on genetic neural network is proposed in this paper, the feasibility and effectiveness of which have been proved by experiments.
Finally, the paper sets forth the author's self-experience in research into signature verification, and it also makes an outlook for further research direction.
引文
[1]willliam K Prratt.邓鲁华,张延恒等译.数字图像处理[M].北京:机械工业出版社,2005.
[2]钟珞,潘昊,封筠,何平主编.模式识别[M].武汉大学出版社,2006年9月.
[3]边肇祺,张学工.编著.模式识别[M].北京:清华大学出版社,2000.
[4]田捷,杨鑫,编著.生物特征识别技术理论与应用,北京:电子工业出版社,2005年9月.
[5]钟珞,饶文碧,邹承明.人工神经网络及其融合应用技术.北京:科学出版社,2007.69-70.
[6]李长河.人工智能及其应用.北京:机械工业出版社,2006.8.228-229.
[7]王小平,曹立明.遗传算法--理论、应用与软件实现.西安:西安交通大学出版社,2002.
[8]R.Plamond,G.Lorette.Automatic signature verification and writer identification-the state of art[J],Pattern Recognition,1989,22(2):107-131.
[9]R.Plamondon.The design of an on-line signature verification system:From theory to practice[J],International Journal of Pattern Recognition and Artificial Intelligence,1994,8(3):795-811.
[10]R.Sabourin,R,Plamondon,L.Beaumier.Structural interpretation of handwritten signature images,International Journal of Pattern Recognition and Artificial Intelligence,1989,22:709-748.
[11]L.Yang,B.K.widjaja,R.Prasead.Application of hidden markov models for signature verification,Pattern Recognition,1995,28:167-170.
[12]H,Cardot,M.Ravenu.A static signature verification system based on a cooperating neural networks architecture,International Journal of Pattern Recognition and Artificial Interlligence,1994,8:679-692.
[13]朱佳鑫,徐全生,龚艺.一种基于游程长度的签名鉴别方法.沈阳工业大学学报,2005,增刊.
[14]赵永健,李永勃.基于小波变换的汉字签名鉴别.计算机工程,2003,29(2):243-245.
[15]求实科技.Visual C++数字图像处理典型算法及实现[M].北京:人民邮电出版社,2006.1-2.
[16]张忻中.汉字识别技术[M].北京:清华大学出版社,1992.
[17]朱浩悦.脱机中文签名鉴定系统关键技术研究:[硕士学位论文].西安:西北大学,2006.
[18]方敏,徐俊艳,王建平等.一种新的文本图像二值化方法.合肥工业大学学报(自然科学版).2001,24(2):166-169.
[19]Qivind D,Torfinn T.Evaluation of binarization methods for document images[J].IEEE Trans PAMI,1995,17(3):312-315.
[20]夏良正.数字图像处理[M].东南大学出版社,2004:150-151.
[21]崔屹.数字图像处理技术与应用.北京:电子工业出版社,1997.
[22]李国旗.脱机中文签名鉴定中的小波分析及识别策略:[硕士学位论文].山东:山东大学,2002.
[23]刘云飞.脱机手写体汉字识别中细化、特征提取和相似字识别算法研究:[湖南大学硕士研究生学位论文].湖南:湖南大学,2006.
[24]孙星明,杨茂江,刘国华等.完全基于结构知识的汉字笔画提取算法.计算机研究与发展,2000,37(5):543-550.
[25]张宏林.Visual C++数字图像模式识别技术及工程实践.北京:人民邮电大学出版社,2003.235-240.
[26]GONZALEZ R C,WOODS R E.数字图像处理[M].阮秋琪,阮宇智,译.北京:电子工业出版社,2003.
[27]王蒙,指纹图像细化算法的改进算法[J].科技信息,2007年第11期.
[28]H.Ogava,K.Taniguchi.Thinning and stroke segmentation for handwritten Chinese character recognition.Pattern Recoginition,1987,1(3):59-68.
[29]M.V.Nagendraprasad,P.S.P Wang,A.Gupta.An improved algorithm for thinning binary digital patterns.In Proceedings of 11th International Conference on Pattern Recognition,The Netherlands Press,1992,386-389.
[30]Roland T.Chin,Hong-Khoon Wan.An one-pass thinning algorithm and ites parallel implementation.Computer Vision,Graphics and Image Processing,1987,30-40.
[31]Richard W.hall.Optimally small operator supports for fully parallel thinning algorithms.IEEE transactions on Pattern Analysis Machine Intelligence.1993,8(15).
[32]黄铁英.手写体汉字识别技术的研究:[硕士学位论文].西安:西安电子科技大学,2004.6-12.
[33]吴佑寿,丁晓青.汉字识别原理、方法与实现.北京:高等教育出版社,1992.
[34]Christophe Lohou,Gilles Bertrand.Two symmetrical thinning algorithms for 3D binary images,based on P-simple points.Pattern Recognition,2007,40(8):2301-2314.
[35]包建军,樊菁.鲁棒的二值图像并行细化算法[J].计算机辅助工程,2006,15(4):1-4.
[36]黄铁英,姜昱明.一种快速手写汉字细化算法[J].计算机工程,2004,30(19):1-3.
[37]曹灿,黄福英.数字图像细化算法研究与实现[J].广西物理,2006,27(3):1-3.
[38]Espinosa-Duro,V.Fingerprints thinning algorithm.Aerospace and Electronic Systems Magazine,IEEE,2003,18(9):28-30.
[39]刘峡壁,贾云得.一种字符图像线段提取及细化算法[J].中国图象图形学报,2005,10(1):48-53.
[40]赵建平,车丹.手写体笔迹单笔划提取算法.长春理工大学学报[J].2005,28(4):66-70.
[41]Vladimir Pervouchine,Graham Leedham.Document Examiner Feature Extraction:Thinned vs.Skeletonised Handwriting Images[J].IEEE,2005,1-6.
[42]李国宏,施鹏飞.基于笔划方向特征和非对称分布的手写体汉字识别[J].上海交通大学学报,2005,39(12):1988-1992.
[43]Ye Wang,Shang-Teng Huang.Chinese word segmentation based on A-priori and adjacent characters[J].Machine Learning and Cybernetics,2005, 6:3808-3813.
[44]Tang,L.Kender,J.R.A off-line stroke-based handwritten word segmentation and recognition method for low-quality educational videos[J].Multimedia Software Engineering,2004,565-568.
[45]Feng Lin,Xiaoou Tang.Dynamic Stroke Information Analysis for Video-Based Handwritten Chinese Character Recognition[J].IEEE,2003,2:1-6.
[46]Lin F,Tang X.Offline handwritten Chinese character stroke extraction[J].In:Proceedings of 16th International Conference on Pattern Recogntion.IEEE,2002,3:249-252.
[47]Hairong Lv,Wenyuan Wang,Chong Wang,Qing Zhuo.Off-line Chinese signature verification based on support vector machines[J].Pattern Recognition Letters,2005,26(15):2390-2399.
[48]袁杰.计算机离线签名鉴定系统研究:[硕士学位论文].新疆:新疆大学,2006.
[49]M.Ammar,Y.Yoshida,T.FuKumura.Feature extraction and selection for simulated Signature Verification.Computer recognition and human production of handwriting[M].USA:Addison-Wesley Co.1989:61-76.
[50]黄戈祥,陈继荣.ART2神经网络在手写体汉字识别中的应用.计算机仿真,2006,23(7).
[51]汪林娟,郑建彬.BP神经网络分类器在签名认证系统中的应用.计算机与数字工程,2007,35(11).
[52]陈静,穆志纯,方新,杜大鹏.基于自组织神经网络的汉字识别模拟研究.计算机工程,2007,33(11).
[53]郭招球,赵跃龙,高敬欣.基于小波和神经网络的车牌字符识别新方法.计算机测量与控制,2006,14(9).
[54]刘明贵,彭俊伟,岳向红,杨永波.基于改进遗传算法的基桩缺陷自动识别.岩土力学,2007,28(10).
[55]钱光超,贾瑞玉,张然,李龙澍.基于一种改进的遗传算法的空间聚类分析.计算机技术与发展,2007,17(12).
[56]曹春红,张斌,李文辉.基于牛顿-遗传混合算法的几何约束问题的求解.系统仿真学报,2007,19(16).
[57]姜丽萍,杜修力.基于经验遗传-单纯形算法和结构模态参数识别结构物理参数的方法.地震工程与工程振动,2007,27(4).
[58]薛迎春,须文波,孙俊.基于遗传模拟退火混合算法的矩形包络求解.计算机工程与设计,2007,28(22).
[59]高鹏东,彭翔,李阿蒙.基于混合遗传算法和点面距离测度的深度像配准.计算机应用研究,2007,24(12).
[60]http://hi.baidu,com/lovelink/blog/item/17805ba955efb5ff1f17a2f7,ht ml
[61]徐丽娜编著.神经网络控制.北京:电子工业出版社,2003.203-203.
[2]钟珞,潘昊,封筠,何平主编.模式识别[M].武汉大学出版社,2006年9月.
[3]边肇祺,张学工.编著.模式识别[M].北京:清华大学出版社,2000.
[4]田捷,杨鑫,编著.生物特征识别技术理论与应用,北京:电子工业出版社,2005年9月.
[5]钟珞,饶文碧,邹承明.人工神经网络及其融合应用技术.北京:科学出版社,2007.69-70.
[6]李长河.人工智能及其应用.北京:机械工业出版社,2006.8.228-229.
[7]王小平,曹立明.遗传算法--理论、应用与软件实现.西安:西安交通大学出版社,2002.
[8]R.Plamond,G.Lorette.Automatic signature verification and writer identification-the state of art[J],Pattern Recognition,1989,22(2):107-131.
[9]R.Plamondon.The design of an on-line signature verification system:From theory to practice[J],International Journal of Pattern Recognition and Artificial Intelligence,1994,8(3):795-811.
[10]R.Sabourin,R,Plamondon,L.Beaumier.Structural interpretation of handwritten signature images,International Journal of Pattern Recognition and Artificial Intelligence,1989,22:709-748.
[11]L.Yang,B.K.widjaja,R.Prasead.Application of hidden markov models for signature verification,Pattern Recognition,1995,28:167-170.
[12]H,Cardot,M.Ravenu.A static signature verification system based on a cooperating neural networks architecture,International Journal of Pattern Recognition and Artificial Interlligence,1994,8:679-692.
[13]朱佳鑫,徐全生,龚艺.一种基于游程长度的签名鉴别方法.沈阳工业大学学报,2005,增刊.
[14]赵永健,李永勃.基于小波变换的汉字签名鉴别.计算机工程,2003,29(2):243-245.
[15]求实科技.Visual C++数字图像处理典型算法及实现[M].北京:人民邮电出版社,2006.1-2.
[16]张忻中.汉字识别技术[M].北京:清华大学出版社,1992.
[17]朱浩悦.脱机中文签名鉴定系统关键技术研究:[硕士学位论文].西安:西北大学,2006.
[18]方敏,徐俊艳,王建平等.一种新的文本图像二值化方法.合肥工业大学学报(自然科学版).2001,24(2):166-169.
[19]Qivind D,Torfinn T.Evaluation of binarization methods for document images[J].IEEE Trans PAMI,1995,17(3):312-315.
[20]夏良正.数字图像处理[M].东南大学出版社,2004:150-151.
[21]崔屹.数字图像处理技术与应用.北京:电子工业出版社,1997.
[22]李国旗.脱机中文签名鉴定中的小波分析及识别策略:[硕士学位论文].山东:山东大学,2002.
[23]刘云飞.脱机手写体汉字识别中细化、特征提取和相似字识别算法研究:[湖南大学硕士研究生学位论文].湖南:湖南大学,2006.
[24]孙星明,杨茂江,刘国华等.完全基于结构知识的汉字笔画提取算法.计算机研究与发展,2000,37(5):543-550.
[25]张宏林.Visual C++数字图像模式识别技术及工程实践.北京:人民邮电大学出版社,2003.235-240.
[26]GONZALEZ R C,WOODS R E.数字图像处理[M].阮秋琪,阮宇智,译.北京:电子工业出版社,2003.
[27]王蒙,指纹图像细化算法的改进算法[J].科技信息,2007年第11期.
[28]H.Ogava,K.Taniguchi.Thinning and stroke segmentation for handwritten Chinese character recognition.Pattern Recoginition,1987,1(3):59-68.
[29]M.V.Nagendraprasad,P.S.P Wang,A.Gupta.An improved algorithm for thinning binary digital patterns.In Proceedings of 11th International Conference on Pattern Recognition,The Netherlands Press,1992,386-389.
[30]Roland T.Chin,Hong-Khoon Wan.An one-pass thinning algorithm and ites parallel implementation.Computer Vision,Graphics and Image Processing,1987,30-40.
[31]Richard W.hall.Optimally small operator supports for fully parallel thinning algorithms.IEEE transactions on Pattern Analysis Machine Intelligence.1993,8(15).
[32]黄铁英.手写体汉字识别技术的研究:[硕士学位论文].西安:西安电子科技大学,2004.6-12.
[33]吴佑寿,丁晓青.汉字识别原理、方法与实现.北京:高等教育出版社,1992.
[34]Christophe Lohou,Gilles Bertrand.Two symmetrical thinning algorithms for 3D binary images,based on P-simple points.Pattern Recognition,2007,40(8):2301-2314.
[35]包建军,樊菁.鲁棒的二值图像并行细化算法[J].计算机辅助工程,2006,15(4):1-4.
[36]黄铁英,姜昱明.一种快速手写汉字细化算法[J].计算机工程,2004,30(19):1-3.
[37]曹灿,黄福英.数字图像细化算法研究与实现[J].广西物理,2006,27(3):1-3.
[38]Espinosa-Duro,V.Fingerprints thinning algorithm.Aerospace and Electronic Systems Magazine,IEEE,2003,18(9):28-30.
[39]刘峡壁,贾云得.一种字符图像线段提取及细化算法[J].中国图象图形学报,2005,10(1):48-53.
[40]赵建平,车丹.手写体笔迹单笔划提取算法.长春理工大学学报[J].2005,28(4):66-70.
[41]Vladimir Pervouchine,Graham Leedham.Document Examiner Feature Extraction:Thinned vs.Skeletonised Handwriting Images[J].IEEE,2005,1-6.
[42]李国宏,施鹏飞.基于笔划方向特征和非对称分布的手写体汉字识别[J].上海交通大学学报,2005,39(12):1988-1992.
[43]Ye Wang,Shang-Teng Huang.Chinese word segmentation based on A-priori and adjacent characters[J].Machine Learning and Cybernetics,2005, 6:3808-3813.
[44]Tang,L.Kender,J.R.A off-line stroke-based handwritten word segmentation and recognition method for low-quality educational videos[J].Multimedia Software Engineering,2004,565-568.
[45]Feng Lin,Xiaoou Tang.Dynamic Stroke Information Analysis for Video-Based Handwritten Chinese Character Recognition[J].IEEE,2003,2:1-6.
[46]Lin F,Tang X.Offline handwritten Chinese character stroke extraction[J].In:Proceedings of 16th International Conference on Pattern Recogntion.IEEE,2002,3:249-252.
[47]Hairong Lv,Wenyuan Wang,Chong Wang,Qing Zhuo.Off-line Chinese signature verification based on support vector machines[J].Pattern Recognition Letters,2005,26(15):2390-2399.
[48]袁杰.计算机离线签名鉴定系统研究:[硕士学位论文].新疆:新疆大学,2006.
[49]M.Ammar,Y.Yoshida,T.FuKumura.Feature extraction and selection for simulated Signature Verification.Computer recognition and human production of handwriting[M].USA:Addison-Wesley Co.1989:61-76.
[50]黄戈祥,陈继荣.ART2神经网络在手写体汉字识别中的应用.计算机仿真,2006,23(7).
[51]汪林娟,郑建彬.BP神经网络分类器在签名认证系统中的应用.计算机与数字工程,2007,35(11).
[52]陈静,穆志纯,方新,杜大鹏.基于自组织神经网络的汉字识别模拟研究.计算机工程,2007,33(11).
[53]郭招球,赵跃龙,高敬欣.基于小波和神经网络的车牌字符识别新方法.计算机测量与控制,2006,14(9).
[54]刘明贵,彭俊伟,岳向红,杨永波.基于改进遗传算法的基桩缺陷自动识别.岩土力学,2007,28(10).
[55]钱光超,贾瑞玉,张然,李龙澍.基于一种改进的遗传算法的空间聚类分析.计算机技术与发展,2007,17(12).
[56]曹春红,张斌,李文辉.基于牛顿-遗传混合算法的几何约束问题的求解.系统仿真学报,2007,19(16).
[57]姜丽萍,杜修力.基于经验遗传-单纯形算法和结构模态参数识别结构物理参数的方法.地震工程与工程振动,2007,27(4).
[58]薛迎春,须文波,孙俊.基于遗传模拟退火混合算法的矩形包络求解.计算机工程与设计,2007,28(22).
[59]高鹏东,彭翔,李阿蒙.基于混合遗传算法和点面距离测度的深度像配准.计算机应用研究,2007,24(12).
[60]http://hi.baidu,com/lovelink/blog/item/17805ba955efb5ff1f17a2f7,ht ml
[61]徐丽娜编著.神经网络控制.北京:电子工业出版社,2003.203-203.