数字识别及其应用
|
摘要
数字识别有着广泛的应用前景。本文介绍了数字识别研究的实际背景、理论意义,以及本文的研究工作等,介绍了两种典型的数字识别算法:基于BP神经网络的识别算法和基于结构特征的识别算法,讨论了印刷体数字识别和基于手写数字识别的主要应用。
对于印刷体数字识别,讨论了一个具体的应用例子:扫描等高线图纸上的数字识别,并提出了一种能够有效地提取扫描等高线图纸中数字并进一步识别的方法。等高线图纸中包含大量的数字注记,正确提取并识别这些数字是图纸矢量化处理中的重要组成部分。本文先用改进的模板匹配的方法提取并初步识别其中的数字,再用特征提取法进一步识别不同的数字。实验结果表明,该算法速度较快、精度高,且有一定的抗干扰性。对于手写体数字识别,本文给出了一个具体的应用例子:学生成绩单中的数字识别。文中对手写体数字的提取及识别的基本原理和方法作了介绍,并用MATLAB工具实现了手写体数字识别系统的主要功能。本文先用MATLAB的函数提取出成绩单中的数字,然后选取数字的结构特征和笔划特征并采用模板匹配的方法达到识别数字的目的,实验结果表明,基于所用结构模型和知识库的识别方法对规范手写体数字是可行的,具有较高的识别率及较好的抗噪性能,也可以识别一定条件下的自由手写体数字。为了提高识别率和可靠性,除了增强对噪声的滤除能力外,还要增大知识库,这些都有待我们进一步的研究。
Digital recognition has the extensively applied foreground. In this paper, we introduce the actual background and the theory meaning and the research work on digital recognition, we introduce two typical algorithms for digital recognition: the recognition algorithm based on BP neural network and the recognition algorithm based on structure characteristic, we discuss the main application on printed digital recognition and handwriting digital recognition.
A specific applied example is discussed for printed digital recognition: digital recognition for the scanned contour drawing. And we put forward a method that can extract the numbers in the drawing efficiently and further recognize these numbers. It includes a great deal of numeral notes in the contour drawing, extracting correctly and recognizing these numbers is an important part in the drawing vector processing. In this paper, we present an advanced method of template matching to extract and recognize the numbers, and then we use the feature extraction method to further recognize different numbers. Experiments show that the speed of the algorithm is quick and the accuracy is high and it has certain anti-interference. A specific applied example is raised for handwriting digital recognition: digital recognition in student's report card. We introduce the basic principle and methods for the handwriting digital extracting and recognition. And we realize the main function of handwriting digital recognition system using the MATLAB. We extract the numbers in the report card using the function of MATLAB, then we select the structure characteristic and the stroke characteristic of the numbers and we present the method of template matching to attain the purpose of recognizing numbers, experiments show that the method based on structure model and knowledge bases is viable for the normative handwriting digital, it also can recognize free handwriting digital under certain condition. In order to enhance recognition rate and reliability, in addition to strengthen the ability of filtering, we still need to enlarge the knowledge base, these all need our further research.
对于印刷体数字识别,讨论了一个具体的应用例子:扫描等高线图纸上的数字识别,并提出了一种能够有效地提取扫描等高线图纸中数字并进一步识别的方法。等高线图纸中包含大量的数字注记,正确提取并识别这些数字是图纸矢量化处理中的重要组成部分。本文先用改进的模板匹配的方法提取并初步识别其中的数字,再用特征提取法进一步识别不同的数字。实验结果表明,该算法速度较快、精度高,且有一定的抗干扰性。对于手写体数字识别,本文给出了一个具体的应用例子:学生成绩单中的数字识别。文中对手写体数字的提取及识别的基本原理和方法作了介绍,并用MATLAB工具实现了手写体数字识别系统的主要功能。本文先用MATLAB的函数提取出成绩单中的数字,然后选取数字的结构特征和笔划特征并采用模板匹配的方法达到识别数字的目的,实验结果表明,基于所用结构模型和知识库的识别方法对规范手写体数字是可行的,具有较高的识别率及较好的抗噪性能,也可以识别一定条件下的自由手写体数字。为了提高识别率和可靠性,除了增强对噪声的滤除能力外,还要增大知识库,这些都有待我们进一步的研究。
Digital recognition has the extensively applied foreground. In this paper, we introduce the actual background and the theory meaning and the research work on digital recognition, we introduce two typical algorithms for digital recognition: the recognition algorithm based on BP neural network and the recognition algorithm based on structure characteristic, we discuss the main application on printed digital recognition and handwriting digital recognition.
A specific applied example is discussed for printed digital recognition: digital recognition for the scanned contour drawing. And we put forward a method that can extract the numbers in the drawing efficiently and further recognize these numbers. It includes a great deal of numeral notes in the contour drawing, extracting correctly and recognizing these numbers is an important part in the drawing vector processing. In this paper, we present an advanced method of template matching to extract and recognize the numbers, and then we use the feature extraction method to further recognize different numbers. Experiments show that the speed of the algorithm is quick and the accuracy is high and it has certain anti-interference. A specific applied example is raised for handwriting digital recognition: digital recognition in student's report card. We introduce the basic principle and methods for the handwriting digital extracting and recognition. And we realize the main function of handwriting digital recognition system using the MATLAB. We extract the numbers in the report card using the function of MATLAB, then we select the structure characteristic and the stroke characteristic of the numbers and we present the method of template matching to attain the purpose of recognizing numbers, experiments show that the method based on structure model and knowledge bases is viable for the normative handwriting digital, it also can recognize free handwriting digital under certain condition. In order to enhance recognition rate and reliability, in addition to strengthen the ability of filtering, we still need to enlarge the knowledge base, these all need our further research.
引文
[1] 王永乾, 吕蓉. 基于BP网络的手写体数字识别方法[J].山东电子, 2004,3: 24～26.
[2] 黄心晔, 王茂祥, 富煜清. 基于结构分析的手写体数字识别算法[J].电子工程师,1999,11: 23～25.
[3] 林晓帆, 丁晓青, 吴佑寿. 手写数字识别技术的应用[J]. http://media.ccidnet.com/media/ciw/663/01450001.htm
[4] 飞思科技产品研发中心. MATLAB 6.5辅助图像处理[M]. 电子工业出版社,2003.
[5] 高彤, 姜华, 吕民. 基于模板匹配的手写体字符识别方法[J]. 哈尔滨工业大学学报,1999,31(1):104～106.
[6] 刘建华, 牛秦洲, 程小辉等. 基于特征的印刷体数字符号识别系统[J]. 桂林工学院学报,2005,25(1):1～3.
[7] 林健. MATLAB 在图像处理与目标识别方面的应用实验[J]. http://www. linjian.cn/
[8] 戴建辉. 自由手写体数字识别系统的设计与实现[J]. http://www.info.wust.edu.cn/download/1/4.doc
[9] 周开利, 康耀红. 神经网络模型及其 MATLAB 仿真程序设计[M]. 北京:清华大学出版社, 2005.
[10] G.Nagy, Spatial Sampling Effects in Optical Character Recognition [J],ICDAR95, pp.309.
[11] Luis R.Blando, Prediction of OCR Accuracy Using Simple Image Features [J], ICDAR95,pp.319.
[12] Kyuisk Chung, Performance Comparison of Several Feature Selection Methods Based on Node Pruning in Handwritten Character Recognition [J],ICDAR97, pp.11.
[13] F.M. Wahl, K.Y.Wong, and R.G.Casey. Block segmentation and text extraction in mixed text/image documents [J]. Computer Graphics and Image Processing, Vol.20,1982,pp375～390.
[14] Jaakko Sauvola and Tapio Seppanen, Adaptive Document Binarization [J], ICDAR97,pp. 147.
[15] A.K.C. Wong, P.K.Sahoo, A gray-level threshold selection method base on maximum entropy principle [J].IEEE Trans. Systems Man and Cybernetics, SMC -19(1989)866～871.
[16] J. Sauvola, M. Pietaksinen, Adaptive document image binarization [J]. Pattern Recognition, 33(2000)225～236.
[17] H.Kamada,K. Fujimoto, High-speed, high-accuracy binarization method for recognizing text in images of low spatial resolution[J].ICDAR99,Int.Conf.On Document Analysis and Recognition, Ulm, Germany,(1999),pp:139～142.
[18] O.D.Trier and A.K.Jain. Goal-directed evaluation of binarization methods[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1995,17(12):1191～1201.
[19] M.Sezgin, B.Sankur, Comparison of thresholding methods for non-destructive testing applications, accepted for IEEE ICIP’2001, International Conference on Image Processing, Thessaloniki, Greece October 7-10,2001.
[20] Victor Wu and R.Manmatha, Text detection technique aids OCR by noise cleanup before binarization [J]. Electronic Imaging(Vol.8,No.2).
[21] L.R.Rabiner. A tutorial on hidden markov models and selected applications in speech recognition [J]. Proceeding of the IEEE, February 1989,77:257～285.
[22] M.I.Sezan, A Peak Detection Algorithm and its Application to Histogram-Based Image Data Reduction[J].Graphical Models and Image Processing,29(1985)47～59.
[23] J.C.Russ, Automatic discrimination of features in gray-scale images[J].Journal of Microscopy,148(3)(1987)263～277.
[24] J.Mao and S.Tomita, An optimal orthonormal system for discriminant analysis [J].Pattern Recognition, vol.18,1985,pp.139～144.
[25] M.Moatar, M.Vahedi. Oneline OCR for handwritten Persion digits: comparison between different feature extraction methods [J].B.Sc project, SECOMP Lab, IUST,1996.
[26] Lei Xu, et al, Methods of Combining Multiple Classifiers and Their Applications to Handwritten Recognition [J]. IEEE Trans. System, Man and Cybernetics, vol. 22, no 3, 1992.6.
[27] C.Y.Suen, et. al. Computer Recognition of Unconstrained Handwritten Numerals [J]. IEEE Trans. on Pattern Anal. And Machine Intell., vol.80, no.7, pp1162～1180.
[28] Lars Kai Hansen, Peter Salamon. Neural Network Ensembles [J].IEEE Tran. on PAMI, vol.12, no.10, pp993～1001, 1990.
[29] Roberto Battiti, Anna Maria Colla. Democracy in Neural Nets: Voting Schemes for Classification [J]. Neural Networks, vol 7, no 4, pp691～707, 1994.
[30] Kagan Tumer, et. al. Analysis of Decision Boundaries in Linearly Combined Neural Classifiers[J]. Pattern Recognition, vol 29, no 2, pp 341～348, 1996.
[31] Roberto Battiti, et. al. Democracy in Neural Nets: Voting Schemes for Classification[J].Neural Networks, vol.7, no.4, pp.691～707, 1994.
[32] 童继进, 刘忠. 基于 Zernike 矩、粗集和神经网络的数字识别方法[J]. 微计算机信息,2005,21(12) :172～174.
[33] 滕书华, 孙即祥, 邵晓芳. 一种鲁棒性的印刷体数字识别算法[J]. 光学与光电技术,2005 ,3 (6) :12～15.
[34] 胡合兴. 经验模型分解在手写体数字识别中的应用[J]. 湖南文理学院学报(自然科学版),2005 ,17 (3): 74～78.
[35] 陈香华, 叶吉祥, 谭冠政. 利用小波和 RBF 神经网络进行手写数字识别[J]. 云南民族大学学报(自然科学版),2005 ,14 (4): 353～355.
[36] 章慎锋, 杨淑莹, 王厚雪. 基于 Bayes 决策的手写体数字识别[J]. 天津理工大学学报,2006 ,22 (1): 78～80.
[37] 洪留荣, 王耀才. 应用图论和基元方向信息的手写数字识别[J]. 计算机工程,2006 ,32 (3) :34～36.
[38] 马立权, 李维, 蔡韩辉等. 手写数字识别中的预处理技术研究[J]. 仪器仪表学报,2001,22(3) :263～265.
[39] 邓丽华, 崔志强. 手写体数字识别系统中一种新的特征提取方案[J]. 湖州师范学院学报,2005 ,27 (2) :99～101.
[40] 苗夺谦, 张红云, 李道国等. 基于主曲线的脱机手写数字识别[J]. 电子学报,2005 ,33 (9) :1639～1643.
[41] 赵万鹏, 古乐野. 基于 Adaboost 的手写体数字识别[J]. 计算机应用,2005, 25 (10) :2413～2415.
[42] 邓丽华, 崔志强, 张静. 基于人工神经网络的手写体数字识别[J]. 三峡大学学报(自然科学版),2005 ,27 (3) :254～256.
[43] 徐大宏, 王润生. 一种实时精确的数字识别方法[J]. 计算机工程与应用,2005 ,41 (2) :87～89.
[44] 苗洪庆, 尹立新, 黄素贞. 基于 DSP 的手写数字识别系统[J]. 计算机工程,2005 ,31 (4) :178～180.
[45] 谭旭, 孔梦荣. 基于层次分解决策树的手写体数字识别[J]. 微电子学与计算机,2004 ,21 (11): 37～39.
[2] 黄心晔, 王茂祥, 富煜清. 基于结构分析的手写体数字识别算法[J].电子工程师,1999,11: 23～25.
[3] 林晓帆, 丁晓青, 吴佑寿. 手写数字识别技术的应用[J]. http://media.ccidnet.com/media/ciw/663/01450001.htm
[4] 飞思科技产品研发中心. MATLAB 6.5辅助图像处理[M]. 电子工业出版社,2003.
[5] 高彤, 姜华, 吕民. 基于模板匹配的手写体字符识别方法[J]. 哈尔滨工业大学学报,1999,31(1):104～106.
[6] 刘建华, 牛秦洲, 程小辉等. 基于特征的印刷体数字符号识别系统[J]. 桂林工学院学报,2005,25(1):1～3.
[7] 林健. MATLAB 在图像处理与目标识别方面的应用实验[J]. http://www. linjian.cn/
[8] 戴建辉. 自由手写体数字识别系统的设计与实现[J]. http://www.info.wust.edu.cn/download/1/4.doc
[9] 周开利, 康耀红. 神经网络模型及其 MATLAB 仿真程序设计[M]. 北京:清华大学出版社, 2005.
[10] G.Nagy, Spatial Sampling Effects in Optical Character Recognition [J],ICDAR95, pp.309.
[11] Luis R.Blando, Prediction of OCR Accuracy Using Simple Image Features [J], ICDAR95,pp.319.
[12] Kyuisk Chung, Performance Comparison of Several Feature Selection Methods Based on Node Pruning in Handwritten Character Recognition [J],ICDAR97, pp.11.
[13] F.M. Wahl, K.Y.Wong, and R.G.Casey. Block segmentation and text extraction in mixed text/image documents [J]. Computer Graphics and Image Processing, Vol.20,1982,pp375～390.
[14] Jaakko Sauvola and Tapio Seppanen, Adaptive Document Binarization [J], ICDAR97,pp. 147.
[15] A.K.C. Wong, P.K.Sahoo, A gray-level threshold selection method base on maximum entropy principle [J].IEEE Trans. Systems Man and Cybernetics, SMC -19(1989)866～871.
[16] J. Sauvola, M. Pietaksinen, Adaptive document image binarization [J]. Pattern Recognition, 33(2000)225～236.
[17] H.Kamada,K. Fujimoto, High-speed, high-accuracy binarization method for recognizing text in images of low spatial resolution[J].ICDAR99,Int.Conf.On Document Analysis and Recognition, Ulm, Germany,(1999),pp:139～142.
[18] O.D.Trier and A.K.Jain. Goal-directed evaluation of binarization methods[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1995,17(12):1191～1201.
[19] M.Sezgin, B.Sankur, Comparison of thresholding methods for non-destructive testing applications, accepted for IEEE ICIP’2001, International Conference on Image Processing, Thessaloniki, Greece October 7-10,2001.
[20] Victor Wu and R.Manmatha, Text detection technique aids OCR by noise cleanup before binarization [J]. Electronic Imaging(Vol.8,No.2).
[21] L.R.Rabiner. A tutorial on hidden markov models and selected applications in speech recognition [J]. Proceeding of the IEEE, February 1989,77:257～285.
[22] M.I.Sezan, A Peak Detection Algorithm and its Application to Histogram-Based Image Data Reduction[J].Graphical Models and Image Processing,29(1985)47～59.
[23] J.C.Russ, Automatic discrimination of features in gray-scale images[J].Journal of Microscopy,148(3)(1987)263～277.
[24] J.Mao and S.Tomita, An optimal orthonormal system for discriminant analysis [J].Pattern Recognition, vol.18,1985,pp.139～144.
[25] M.Moatar, M.Vahedi. Oneline OCR for handwritten Persion digits: comparison between different feature extraction methods [J].B.Sc project, SECOMP Lab, IUST,1996.
[26] Lei Xu, et al, Methods of Combining Multiple Classifiers and Their Applications to Handwritten Recognition [J]. IEEE Trans. System, Man and Cybernetics, vol. 22, no 3, 1992.6.
[27] C.Y.Suen, et. al. Computer Recognition of Unconstrained Handwritten Numerals [J]. IEEE Trans. on Pattern Anal. And Machine Intell., vol.80, no.7, pp1162～1180.
[28] Lars Kai Hansen, Peter Salamon. Neural Network Ensembles [J].IEEE Tran. on PAMI, vol.12, no.10, pp993～1001, 1990.
[29] Roberto Battiti, Anna Maria Colla. Democracy in Neural Nets: Voting Schemes for Classification [J]. Neural Networks, vol 7, no 4, pp691～707, 1994.
[30] Kagan Tumer, et. al. Analysis of Decision Boundaries in Linearly Combined Neural Classifiers[J]. Pattern Recognition, vol 29, no 2, pp 341～348, 1996.
[31] Roberto Battiti, et. al. Democracy in Neural Nets: Voting Schemes for Classification[J].Neural Networks, vol.7, no.4, pp.691～707, 1994.
[32] 童继进, 刘忠. 基于 Zernike 矩、粗集和神经网络的数字识别方法[J]. 微计算机信息,2005,21(12) :172～174.
[33] 滕书华, 孙即祥, 邵晓芳. 一种鲁棒性的印刷体数字识别算法[J]. 光学与光电技术,2005 ,3 (6) :12～15.
[34] 胡合兴. 经验模型分解在手写体数字识别中的应用[J]. 湖南文理学院学报(自然科学版),2005 ,17 (3): 74～78.
[35] 陈香华, 叶吉祥, 谭冠政. 利用小波和 RBF 神经网络进行手写数字识别[J]. 云南民族大学学报(自然科学版),2005 ,14 (4): 353～355.
[36] 章慎锋, 杨淑莹, 王厚雪. 基于 Bayes 决策的手写体数字识别[J]. 天津理工大学学报,2006 ,22 (1): 78～80.
[37] 洪留荣, 王耀才. 应用图论和基元方向信息的手写数字识别[J]. 计算机工程,2006 ,32 (3) :34～36.
[38] 马立权, 李维, 蔡韩辉等. 手写数字识别中的预处理技术研究[J]. 仪器仪表学报,2001,22(3) :263～265.
[39] 邓丽华, 崔志强. 手写体数字识别系统中一种新的特征提取方案[J]. 湖州师范学院学报,2005 ,27 (2) :99～101.
[40] 苗夺谦, 张红云, 李道国等. 基于主曲线的脱机手写数字识别[J]. 电子学报,2005 ,33 (9) :1639～1643.
[41] 赵万鹏, 古乐野. 基于 Adaboost 的手写体数字识别[J]. 计算机应用,2005, 25 (10) :2413～2415.
[42] 邓丽华, 崔志强, 张静. 基于人工神经网络的手写体数字识别[J]. 三峡大学学报(自然科学版),2005 ,27 (3) :254～256.
[43] 徐大宏, 王润生. 一种实时精确的数字识别方法[J]. 计算机工程与应用,2005 ,41 (2) :87～89.
[44] 苗洪庆, 尹立新, 黄素贞. 基于 DSP 的手写数字识别系统[J]. 计算机工程,2005 ,31 (4) :178～180.
[45] 谭旭, 孔梦荣. 基于层次分解决策树的手写体数字识别[J]. 微电子学与计算机,2004 ,21 (11): 37～39.