基于支持向量机的货币识别研究
|
摘要
论文围绕支持向量机分类器的算法及支持向量机应用于货币识别进行了一些探索性的研究。货币识别是一个小样本、非线性和高维模式识别问题，是当前模式识别中的难题之一，具有重要的研究意义和实用价值。支持向量机是一种新的基于统计学习理论的机器学习算法，它可以应用于小样本、非线性和高维模式识别。
训练算法采用一种支持向量机快速算法—基于壳向量的支持向量机算法(HVSVM)。利用训练样本集中的几何信息，在样本中首先选取一部分最有可能成为支持向量的样本—壳向量，将其作为新的训练样本集再进行SVM训练。由于在提取壳向量的过程中只需线性规划运算，之后的训练过程又只需处理原训练样本中的一部分，大大降低了二次规划过程的时间，使整个算法的训练速度大为提高。
多值分类算法采用一种基于核聚类方法的多层次SVM分类树。该算法将核空间中的无监督学习方法和有监督学习方法结合起来，按照样本集逐层核聚类的结果对多层次分类树的子任务进行定义，使其更为准确有效。实现了一种结构更加简洁计算更有效率的多层SVM分类树算法。
研究了支持向量机的学习算法，依据支持向量机的特点采用了对应的货币特征数据获取及预处理方法，提出采用HVSVM和改进的SVM分类树结构的多类分类器构建的支持向量机用于货币识别，从而达到对货币高效、准确识别。实验结果证实了该方案的有效性。
SVM algorithms and currency recognition based on support vector machines are mainly discussed in this paper. Currency recognition which is a scared samples, nonlinear and high dimensions pattern recognition problem is one of the difficult problems of modern pattern recognition and of specific research significance and practical value. The Support Vector Machines is a new machines learning algorithm based on the Statistical Learning Theory. It has been applied in pattern recognition of scared samples, nonlinear and high dimensions character vectors.
A new fast algorithm of support vector machines and the concept of "hullvector" are proposed. Using the geometric information of the training samples, the algorithm first extracts the set of hullvectors, which are most likely to be the support vector. Then the set of hullvectors are trained as the new training samples to get the support vectors. HVSVM reduces the time consumed by the QP problem in the SVM training in large degree, and highly speeds the whole training process of SVM.
A hierarchical Support Vector Machines classify tree based on kernel clustering method combining the unsupervised learning method and supervised learning together is proposed to apply. It proves the algorithm is more effective and simple in structure and performance better than the original algorithm.
The algorithm of Support Vector Machines has been studied in this paper, according to the specialty of Support Vector Machines, corresponding method of obtaining and advance processing currency character data recognition is applied to currency recognition, and the HVSVM training algorithm and hierarchical SVM tree of multi-values classify method is proposed to used. Consequently, efficient and accurate paper currency recognition is achieved, and the results show the validity of the project.
训练算法采用一种支持向量机快速算法—基于壳向量的支持向量机算法(HVSVM)。利用训练样本集中的几何信息，在样本中首先选取一部分最有可能成为支持向量的样本—壳向量，将其作为新的训练样本集再进行SVM训练。由于在提取壳向量的过程中只需线性规划运算，之后的训练过程又只需处理原训练样本中的一部分，大大降低了二次规划过程的时间，使整个算法的训练速度大为提高。
多值分类算法采用一种基于核聚类方法的多层次SVM分类树。该算法将核空间中的无监督学习方法和有监督学习方法结合起来，按照样本集逐层核聚类的结果对多层次分类树的子任务进行定义，使其更为准确有效。实现了一种结构更加简洁计算更有效率的多层SVM分类树算法。
研究了支持向量机的学习算法，依据支持向量机的特点采用了对应的货币特征数据获取及预处理方法，提出采用HVSVM和改进的SVM分类树结构的多类分类器构建的支持向量机用于货币识别，从而达到对货币高效、准确识别。实验结果证实了该方案的有效性。
SVM algorithms and currency recognition based on support vector machines are mainly discussed in this paper. Currency recognition which is a scared samples, nonlinear and high dimensions pattern recognition problem is one of the difficult problems of modern pattern recognition and of specific research significance and practical value. The Support Vector Machines is a new machines learning algorithm based on the Statistical Learning Theory. It has been applied in pattern recognition of scared samples, nonlinear and high dimensions character vectors.
A new fast algorithm of support vector machines and the concept of "hullvector" are proposed. Using the geometric information of the training samples, the algorithm first extracts the set of hullvectors, which are most likely to be the support vector. Then the set of hullvectors are trained as the new training samples to get the support vectors. HVSVM reduces the time consumed by the QP problem in the SVM training in large degree, and highly speeds the whole training process of SVM.
A hierarchical Support Vector Machines classify tree based on kernel clustering method combining the unsupervised learning method and supervised learning together is proposed to apply. It proves the algorithm is more effective and simple in structure and performance better than the original algorithm.
The algorithm of Support Vector Machines has been studied in this paper, according to the specialty of Support Vector Machines, corresponding method of obtaining and advance processing currency character data recognition is applied to currency recognition, and the HVSVM training algorithm and hierarchical SVM tree of multi-values classify method is proposed to used. Consequently, efficient and accurate paper currency recognition is achieved, and the results show the validity of the project.
引文
[1] 湖南省金融货币识别与自助服务平台工程技术研究中心．KCM系列多国货币识别器[R]．银税类嵌入式软件集成开发平台技术白皮书．2005，12：118～126
[2] 湖南省金融货币识别与自助服务平台工程技术研究中心．基于货币识别的智能自助服务平台[R]．国家科技攻关计划引导项目可行性报告．2003
[3] Vapnik V. The Nature of Statistical Learning Theory[M]. New York: Springer Verlag, 1995
[4] Nello Cristianini etc．支持向量机导论[M]．北京：电子工业出版社，2004．
[5] 湖南省金融货币识别与自助服务平台工程技术研究中心．多国货币纸硬币识别算法模块库[R]．湖南省金融货币识别与自助服务平台技术报告，2005
[6] 湖南省金融货币识别与自助服务平台工程技术研究中心．多国货币纸硬币嵌入面额数据特征库与检伪特征数据库[R]．湖南省金融货币识别与自助服务平台技术报告，2005
[7] 刘丁，彭军民等．基于SCG的货币识别研究．中国人工智能学会年会(CAAI-10)论文集．
[8] F. Takeda, S. Omatu, T. Inoue. An Expert System for Determing the Discrimi tive Functions for Bill Money Recognition. Trans. Info. Process. Soc. Japan, 1992, 33(7):80～90
[9] F. Takeda, T. Nishikage, Y. Matsumoto. Characteristics Extraction of Paper Currency using Symmetrical Masks Optimized by GA and Neuro-Recognition of Multi-National Paper. Currency. Proceedings of IEEE World Congress on Computational Intelligence, Alaska, USA, 1998(1):634～639
[10] Jae-Kang Lee, Seong-Goo Jeon, Ⅱ-Hwan Kim Distinctive Point Extraction and Recognition Algorithm for Various Kinds of Euro Banknotes. International Journal of Control, Automation and Systems, 2004, 2(2):201～206
[11] 张平，徐问之．基于神经网络多国货币种类的识别与研究[J]．重庆大学学报(自然科学版)，1999，22(3)：34～40
[12] 殷泽兴，钱浙滨，李介谷．基于数学形态学和神经元网络的货币识别[J]．上海交通大学学报，1999，33(9)：1142～1145
[13] 谢凯，郝建新．一种基于货币特征的神经网络模式识别算法[J]．计算机应用，2002，22(11)：76～77
[14] 刘家锋，刘松波，唐降龙．一种实时纸币识别方法的研究[J]．计算机研究与发展，2003，40(7)：1057～1061
[15] 张颖，陈雪波等．人民币纸币识别系统[J]．仪器仪表学报．2004，25(4)：669～670．
[16] 陈慧鹏，杨亮亮，李鸿，周云飞．模糊集识别法在纸币清分中的应用研究[J]．机床与液压．2005，(6)：167～169
[17] 杨相珀，满庆丰．钞币面值和真伪的快速识别算法的研究[J]．计算机工程与应用．2005，(25)：209～211．
[18] F. Takeda, S. Omatu. High-speed Paper Currency Recognition by Neural Networks. IEEE Trans. on Neural Networks. 1995, 6(1):73～77
[19] Osuna E, Freund R, Girosi F. Training Support Vector Machine: An Application to Face Detection[C].In:Proceedings of IEEE International Conference on Computer Vision and Pattern Recognition, New York: IEEE, 1997:130～136
[20] Dumais S, Platt J etc. Inductive Learning Algorithms and Representations for Text Categorization[C]. In: Proceedings of the 7th International Conference on Information and Knowledge Management, 1998
[21] Joachims T. Text Categorization with Support Vector Machines: Learning with Relevant Features[C]. In:Proceedings of the 10th European Conference on Machine Leaning, 1998
[22] 边肇祺等．模式识别(第二版)[M]．清华大学出版社，2000
[23] Vapnik V．统计学习理论的本质[M]．张学工．北京：清华大学出版社，2000
[24] Cortes C, Vapnik V. Support Vector Networks. Machine Learning[J]. 1995, (20):273～297
[25] Courant R, Hilbert D. Methods of Mathematical Physics[M]. Berlin:Springer-Verlag, 1953
[26] Cui Weidong, Zhou Zhihua, Li Xing. Research of Support Vector Machines[J]. Computer Engineering and Application. 2001, 37(1):86～88.
[27] 白亮，老松杨，胡艳丽．支持向量机训练算法比较研究[J]．计算机工程与应用，2005，41(17)：79～84
[28] 施光燕，董加礼．最优化方法[M]．高等教育出版社，1999
[29] 胡适耕，施保昌．最优化原理[M]．武汉：华中理工大学出版社，2000
[30] Osuna E, Freund R, Girosi F. An Improved Training Algorithm for Support Vector Machines[C]. In:Principe J, Gile L, Morgan N, Wilson eds. Proceedings of the 1997 IEEE Workshop on Neural Networks for Signal Proceeding, New York:IEEE, 1997:276～285
[31] Joachims T. Making Large-Scale SVM Learning Practical[C]. In:Scholkpf B, Burges C J C, Smola A eds. Advances in Kernel Methods Support Vector Learning, Cambridge, MA:MIT Press, 1998:169～184
[32] Chang C C, Hsu C W, Lin C J. The analysis of decomposition methods for support vector machines[C]. In: Workshop on Support Vector machines, IJCAI, 1999
[33] Platt J C. Fast Training of SVMs Using Sequential Minimal Optimization. In: Scholkopf B, Burges C J C, Smola A J eds. Advances in Kernel Methods-Support Vector Learning, Cambridge, MA: MIT Press 1998: 185～208
[34] Keerthi, S. S. etc. Improvements to Platt's SMO algorithm for SVM classier Design[C]. TR CD-99-14, Dept. of Mecha. and Prod. Engin. National Uni. Of Singapore, 1999
[35] Chih W H. ect. A simple decomposition Method for support vector machines. Technical report, National Taiwan university, 1999
[36] Ahmed S N. Incremental Learning with Support Vector Machines (IJCAI 1999)[C]. In: Workshop on Support Vector Machines, Stockholm, Sweden, 1999
[37] Cauwenberghs G, Poggio T. Incremental and Decremental Support Vector Machine[J]. Machine Learning. 2001, 44(13): 409～415
[38] Zhang Xuegong. Using class center vectors to build support vector machines. Neural Networks for Signal Processing[C]. New York: IEEE Press, 1999, 3～11.
[39] Keerthi S, Shevade S, Bhattcharyya C, et al. A fast iterative nearest point algorithm for support vector machine classifier design [J]. IEEE Transactions on Neural Network, 2000,11(1): 124～136
[40] Yang M H, hhuja N. A Geometric Approach to Train Support Vector Machines [C]. In: Proceedings of CVPR 2000, Hilton Head Island. 2000: 430～437
[41] C. Bradford Barber. The quickhull algorithm for convex hulls, ACM Trans. Mathematical Software. 1996, 22(4): 469～483
[42] Y. Freund, R. E. Schapire. Large margin classification using the perceptron algorithm. Machine learning. 1999, 37(3): 277～296
[43] Weston J., Watkin C. Multiclass support vector machines, TR CSDTR9804, Department of Computer Scinece Egham, Surrey TW20 OEX, England, 1998
[44] Platt J. etc. Large margin BAGs for multiclassification, in Advances in Neural information processing systems12, MIT Press. 2000: 547～553.
[45] Zhan Yong, Zhou Yanhong, Lu Zhengding. A new method to improve the sensitivity of support vector machine based on data optimization, Robotics, Intelligent Systems and Signal Processing, IEEE International Conference. 2003, (2): 892～896
[46] Schwenker. F. Hierarchical support vector machines for multi-class pattern recognition, Knowledge-Based Intelligent Engineering Systems and Allied Technologies, 2000. Proceedings. Fourth International Conference. 2000, (2),561～565
[47] Mark Girllami, Mercer Kernel Based Clustering in Feature Space, IEEE Transactions on Neural Networks, 2002,13(3):780～784
[48] 张国宣．基于统计学习理论的支持向量机分类方法研究．[博士学位论文]．合肥：中国科技大学，2004．
[49] Takeda F., Omatu S.,T. Inoue etc. Bill Money Recognition using Neural Network with FFT as Pre-processOr. Trans. Inst. Syst, Control and Info. Eng, Japan, 1992,5(7):265～273
[50] Ahmadi. A., Omatu. S. etc. A Reliable Classification Method for Paper Currency Based on the Non-linear PCA[C]. IEEJ Trans. EIS. 2003, 123(10): 1783～1789
[51] 魏广芬，唐祯安，余隽．基于主成分分析和BP神经网络的气体识别方法研究．传感技术学报，2001，(4)：1～2
[52] Ahmadi. A, Omatu. S. Impiementing a reliable neuro-classifier for paper currency using PCA algorithm. SICE 2002. Proceedings of the 41st SICE Annual Conference. 2002, (4):2466～2468
[53] Qiu. B. Prinet, V. Perrier. E. Image Analysis and Processing. Proceedings of the 12th International Conference. 2003:385～390
[54] Chen. X, FLYnn. P.J etc. K.W. Analysis and Modeling of Faces and Gestures. IEEE International Workshop. 2003:127～134
[55] 孙即祥等．现代模式识别[M]．长沙：国防科技大学出版社，2002．
[56] 薛定宇．科学运算语言MATLAB5.3程序设计与应用[M]．北京：清华大学出版社，2000．
[57] 飞思科技产品研发中心．MATLAB6.5辅助图像处理[M]．北京：电子工业出版社，2003．
[58] 吴旭光等．计算机仿真技术[M]北京：化学工业出版社，2005．
[59] 飞思科技产品研发中心．神经网络理论与MATLAB 7实现[M]．北京：电子工业出版社，2005．
[60] 陈荣淋，林建华等．支持向量机在砂土液化预测中的应用研究[J]．中国地质灾害与防治学报，2005，16(2)：15～18．
[61] 彭军民．基于神经网络的货币识别研究．[硕士学位论文]．西安：西安理工大学，2004．
[2] 湖南省金融货币识别与自助服务平台工程技术研究中心．基于货币识别的智能自助服务平台[R]．国家科技攻关计划引导项目可行性报告．2003
[3] Vapnik V. The Nature of Statistical Learning Theory[M]. New York: Springer Verlag, 1995
[4] Nello Cristianini etc．支持向量机导论[M]．北京：电子工业出版社，2004．
[5] 湖南省金融货币识别与自助服务平台工程技术研究中心．多国货币纸硬币识别算法模块库[R]．湖南省金融货币识别与自助服务平台技术报告，2005
[6] 湖南省金融货币识别与自助服务平台工程技术研究中心．多国货币纸硬币嵌入面额数据特征库与检伪特征数据库[R]．湖南省金融货币识别与自助服务平台技术报告，2005
[7] 刘丁，彭军民等．基于SCG的货币识别研究．中国人工智能学会年会(CAAI-10)论文集．
[8] F. Takeda, S. Omatu, T. Inoue. An Expert System for Determing the Discrimi tive Functions for Bill Money Recognition. Trans. Info. Process. Soc. Japan, 1992, 33(7):80～90
[9] F. Takeda, T. Nishikage, Y. Matsumoto. Characteristics Extraction of Paper Currency using Symmetrical Masks Optimized by GA and Neuro-Recognition of Multi-National Paper. Currency. Proceedings of IEEE World Congress on Computational Intelligence, Alaska, USA, 1998(1):634～639
[10] Jae-Kang Lee, Seong-Goo Jeon, Ⅱ-Hwan Kim Distinctive Point Extraction and Recognition Algorithm for Various Kinds of Euro Banknotes. International Journal of Control, Automation and Systems, 2004, 2(2):201～206
[11] 张平，徐问之．基于神经网络多国货币种类的识别与研究[J]．重庆大学学报(自然科学版)，1999，22(3)：34～40
[12] 殷泽兴，钱浙滨，李介谷．基于数学形态学和神经元网络的货币识别[J]．上海交通大学学报，1999，33(9)：1142～1145
[13] 谢凯，郝建新．一种基于货币特征的神经网络模式识别算法[J]．计算机应用，2002，22(11)：76～77
[14] 刘家锋，刘松波，唐降龙．一种实时纸币识别方法的研究[J]．计算机研究与发展，2003，40(7)：1057～1061
[15] 张颖，陈雪波等．人民币纸币识别系统[J]．仪器仪表学报．2004，25(4)：669～670．
[16] 陈慧鹏，杨亮亮，李鸿，周云飞．模糊集识别法在纸币清分中的应用研究[J]．机床与液压．2005，(6)：167～169
[17] 杨相珀，满庆丰．钞币面值和真伪的快速识别算法的研究[J]．计算机工程与应用．2005，(25)：209～211．
[18] F. Takeda, S. Omatu. High-speed Paper Currency Recognition by Neural Networks. IEEE Trans. on Neural Networks. 1995, 6(1):73～77
[19] Osuna E, Freund R, Girosi F. Training Support Vector Machine: An Application to Face Detection[C].In:Proceedings of IEEE International Conference on Computer Vision and Pattern Recognition, New York: IEEE, 1997:130～136
[20] Dumais S, Platt J etc. Inductive Learning Algorithms and Representations for Text Categorization[C]. In: Proceedings of the 7th International Conference on Information and Knowledge Management, 1998
[21] Joachims T. Text Categorization with Support Vector Machines: Learning with Relevant Features[C]. In:Proceedings of the 10th European Conference on Machine Leaning, 1998
[22] 边肇祺等．模式识别(第二版)[M]．清华大学出版社，2000
[23] Vapnik V．统计学习理论的本质[M]．张学工．北京：清华大学出版社，2000
[24] Cortes C, Vapnik V. Support Vector Networks. Machine Learning[J]. 1995, (20):273～297
[25] Courant R, Hilbert D. Methods of Mathematical Physics[M]. Berlin:Springer-Verlag, 1953
[26] Cui Weidong, Zhou Zhihua, Li Xing. Research of Support Vector Machines[J]. Computer Engineering and Application. 2001, 37(1):86～88.
[27] 白亮，老松杨，胡艳丽．支持向量机训练算法比较研究[J]．计算机工程与应用，2005，41(17)：79～84
[28] 施光燕，董加礼．最优化方法[M]．高等教育出版社，1999
[29] 胡适耕，施保昌．最优化原理[M]．武汉：华中理工大学出版社，2000
[30] Osuna E, Freund R, Girosi F. An Improved Training Algorithm for Support Vector Machines[C]. In:Principe J, Gile L, Morgan N, Wilson eds. Proceedings of the 1997 IEEE Workshop on Neural Networks for Signal Proceeding, New York:IEEE, 1997:276～285
[31] Joachims T. Making Large-Scale SVM Learning Practical[C]. In:Scholkpf B, Burges C J C, Smola A eds. Advances in Kernel Methods Support Vector Learning, Cambridge, MA:MIT Press, 1998:169～184
[32] Chang C C, Hsu C W, Lin C J. The analysis of decomposition methods for support vector machines[C]. In: Workshop on Support Vector machines, IJCAI, 1999
[33] Platt J C. Fast Training of SVMs Using Sequential Minimal Optimization. In: Scholkopf B, Burges C J C, Smola A J eds. Advances in Kernel Methods-Support Vector Learning, Cambridge, MA: MIT Press 1998: 185～208
[34] Keerthi, S. S. etc. Improvements to Platt's SMO algorithm for SVM classier Design[C]. TR CD-99-14, Dept. of Mecha. and Prod. Engin. National Uni. Of Singapore, 1999
[35] Chih W H. ect. A simple decomposition Method for support vector machines. Technical report, National Taiwan university, 1999
[36] Ahmed S N. Incremental Learning with Support Vector Machines (IJCAI 1999)[C]. In: Workshop on Support Vector Machines, Stockholm, Sweden, 1999
[37] Cauwenberghs G, Poggio T. Incremental and Decremental Support Vector Machine[J]. Machine Learning. 2001, 44(13): 409～415
[38] Zhang Xuegong. Using class center vectors to build support vector machines. Neural Networks for Signal Processing[C]. New York: IEEE Press, 1999, 3～11.
[39] Keerthi S, Shevade S, Bhattcharyya C, et al. A fast iterative nearest point algorithm for support vector machine classifier design [J]. IEEE Transactions on Neural Network, 2000,11(1): 124～136
[40] Yang M H, hhuja N. A Geometric Approach to Train Support Vector Machines [C]. In: Proceedings of CVPR 2000, Hilton Head Island. 2000: 430～437
[41] C. Bradford Barber. The quickhull algorithm for convex hulls, ACM Trans. Mathematical Software. 1996, 22(4): 469～483
[42] Y. Freund, R. E. Schapire. Large margin classification using the perceptron algorithm. Machine learning. 1999, 37(3): 277～296
[43] Weston J., Watkin C. Multiclass support vector machines, TR CSDTR9804, Department of Computer Scinece Egham, Surrey TW20 OEX, England, 1998
[44] Platt J. etc. Large margin BAGs for multiclassification, in Advances in Neural information processing systems12, MIT Press. 2000: 547～553.
[45] Zhan Yong, Zhou Yanhong, Lu Zhengding. A new method to improve the sensitivity of support vector machine based on data optimization, Robotics, Intelligent Systems and Signal Processing, IEEE International Conference. 2003, (2): 892～896
[46] Schwenker. F. Hierarchical support vector machines for multi-class pattern recognition, Knowledge-Based Intelligent Engineering Systems and Allied Technologies, 2000. Proceedings. Fourth International Conference. 2000, (2),561～565
[47] Mark Girllami, Mercer Kernel Based Clustering in Feature Space, IEEE Transactions on Neural Networks, 2002,13(3):780～784
[48] 张国宣．基于统计学习理论的支持向量机分类方法研究．[博士学位论文]．合肥：中国科技大学，2004．
[49] Takeda F., Omatu S.,T. Inoue etc. Bill Money Recognition using Neural Network with FFT as Pre-processOr. Trans. Inst. Syst, Control and Info. Eng, Japan, 1992,5(7):265～273
[50] Ahmadi. A., Omatu. S. etc. A Reliable Classification Method for Paper Currency Based on the Non-linear PCA[C]. IEEJ Trans. EIS. 2003, 123(10): 1783～1789
[51] 魏广芬，唐祯安，余隽．基于主成分分析和BP神经网络的气体识别方法研究．传感技术学报，2001，(4)：1～2
[52] Ahmadi. A, Omatu. S. Impiementing a reliable neuro-classifier for paper currency using PCA algorithm. SICE 2002. Proceedings of the 41st SICE Annual Conference. 2002, (4):2466～2468
[53] Qiu. B. Prinet, V. Perrier. E. Image Analysis and Processing. Proceedings of the 12th International Conference. 2003:385～390
[54] Chen. X, FLYnn. P.J etc. K.W. Analysis and Modeling of Faces and Gestures. IEEE International Workshop. 2003:127～134
[55] 孙即祥等．现代模式识别[M]．长沙：国防科技大学出版社，2002．
[56] 薛定宇．科学运算语言MATLAB5.3程序设计与应用[M]．北京：清华大学出版社，2000．
[57] 飞思科技产品研发中心．MATLAB6.5辅助图像处理[M]．北京：电子工业出版社，2003．
[58] 吴旭光等．计算机仿真技术[M]北京：化学工业出版社，2005．
[59] 飞思科技产品研发中心．神经网络理论与MATLAB 7实现[M]．北京：电子工业出版社，2005．
[60] 陈荣淋，林建华等．支持向量机在砂土液化预测中的应用研究[J]．中国地质灾害与防治学报，2005，16(2)：15～18．
[61] 彭军民．基于神经网络的货币识别研究．[硕士学位论文]．西安：西安理工大学，2004．