面向多媒体学习的汉字图像识别技术研究
|
摘要
文字图像识别技术是一种运用计算机实现自动辨别图像文字的实用技术。它广泛应用于不同领域。文字图像识别技术用于幼儿教育,能够为幼儿提供一种新的人机交互形式。另外,多媒体技术可以为幼儿的汉字学习提供一个图、文、声并茂的学习环境,但是缺乏处理真实世界信息的能力。因此,文字图像识别技术和多媒体技术结合可以形成一种面向多媒体学习的汉字图像识别技术。它不仅可以实现对汉字图像的自动识别,而且可以把汉字图像以多媒体信息如图片、声音、视频、音频、动画等形式表现出来,弥补汉字在传统教育过程中死记硬背的缺点,降低汉字学习的难度,提高学生的动手能力,促使幼儿更加积极、主动、认真、专注的学习,为幼儿的汉字教育提供一条更加有效的途径。
本文在对汉字图像识别技术研究的基础上,结合多媒体信息,实现了对汉字图像的识别和多媒体信息的表示。然后运用这些技术开发一个汉字学习的原型系统。系统经过测试,证明了本文算法不仅能够有效的完成汉字图像识别,并且能以多媒体信息的形式辅助学习。
论文的主要工作如下:
(1)特征选取以及分类器的设计。根据汉字特点,提取笔画密度和灰度投影作为汉字特征。根据不同的笔画范围采用动态的特征提取法。然后根据提取特征范围选择不同的分类器进行汉字识别。针对灰度投影特征,本文采用灰度投影模板匹配的方法实现对汉字图像的识别。
(2)冲突处理机制。针对汉字图像识别过程中出现的汉字冲突问题,运用冲突处理机制来处理。冲突主要分两种情况:笔画密度冲突和灰度投影冲突。笔画密度冲突主要分两种:一字多码和一码多字。根据出现的情况不同,采取不同的冲突处理机制。
(3)学习机制。针对汉字识别过程中运用冲突处理机制仍然不能实现汉字识别的情况,采取学习机制对其进行学习,完成对汉字特征的提取以及多媒体信息的输入。根据学习过程中特征提取的不同,学习机制可以分为两类:笔画密度学习机制和灰度投影学习机制。并且针对在模板匹配过程中出现的阈值选择问题,采用学习机制实现对阈值的自动确定。
The technology to recognize Chinese character in image is a practical technology that can automatically recognize the Chinese character by using computers. It is widely used in different areas. The character recognition technology can also be applied to childhood education, which provides the young children a new human-computer interaction form .In additional, the multimedia technology can provide the children a good Chinese character learning environment full of images, text and sound. Thus the character recognition technology and multimedia technology can combine to form the Chinese characters recognition technology particularly for multi-media learning. It can not only identify the character image automatically, but also it can present the character images in various forms of multimedia information such as pictures, sounds, video, audio, animation and so on, which can provide children a good learning environment full of images, text and sound. It makes up the shortcomings like the emptiness and uncertainty of the shape and meaning of Chinese characters in traditional education, and also the shortcomings of rote learning. In addition, it can reduce the difficulty of Chinese characters learning, enhance the practical ability of students, and promote children to learn more actively, seriously and attentively, which in all provide a more effective way for children to learn Chinese characters.
Based on the research of the Chinese character image recognition technology, and combined with multimedia information, this paper realizes the representation of the character image recognition and multimedia information. Then, these technologies are used to develop a system prototype for Chinese character learning. And the algorithm in this paper proves to be effective after the system test.
The main contents are as follows:
(1) Feature selection and the design of classifier. Based on the features of Chinese characters, the density of stroke and the gray-scale projection are extracted to be the features of Chinese characters. And the method of feature extraction by dynamic classification is applied according to the different stroke scopes. Then, different classifiers are selectively designed according to the different scopes of feature extraction. Based on the feature of the gray-scale projection, this paper introduces the method of template matching of gray-scale projection to realize the Chinese character image recognition.
(2) The conflict management mechanism. The conflict management mechanism is used to solve the problem of Chinese characters conflicts that appear in the process of Chinese characters image recognition. The conflicts mainly can be divided into two cases: multi-codes for sole-character and multi-characters for sole-code and. According to different scenarios, this paper will adopt different conflict management mechanism.
(3)The learning mechanism. To solve the problem that some Chinese characters can't be identified in the process of Chinese character recognition, this paper uses the learning mechanism to extract the characteristics of Chinese characters and input the multimedia information. According to the difference of the extracted feature during the learning process, the learning mechanism can be divided into kinds that are the stroke-density learning mechanism and the gray-scale projection learning mechanism. Meanwhile, As to the problem of threshold choice in the process of template matching, this paper chooses the corresponding learning mechanism to determine the threshold automatically.
At last this paper develops the system prototype based on the above algorithm and the results prove that the system can effectively identify the image of Chinese characters, and can aid children's learning by multimedia information.
本文在对汉字图像识别技术研究的基础上,结合多媒体信息,实现了对汉字图像的识别和多媒体信息的表示。然后运用这些技术开发一个汉字学习的原型系统。系统经过测试,证明了本文算法不仅能够有效的完成汉字图像识别,并且能以多媒体信息的形式辅助学习。
论文的主要工作如下:
(1)特征选取以及分类器的设计。根据汉字特点,提取笔画密度和灰度投影作为汉字特征。根据不同的笔画范围采用动态的特征提取法。然后根据提取特征范围选择不同的分类器进行汉字识别。针对灰度投影特征,本文采用灰度投影模板匹配的方法实现对汉字图像的识别。
(2)冲突处理机制。针对汉字图像识别过程中出现的汉字冲突问题,运用冲突处理机制来处理。冲突主要分两种情况:笔画密度冲突和灰度投影冲突。笔画密度冲突主要分两种:一字多码和一码多字。根据出现的情况不同,采取不同的冲突处理机制。
(3)学习机制。针对汉字识别过程中运用冲突处理机制仍然不能实现汉字识别的情况,采取学习机制对其进行学习,完成对汉字特征的提取以及多媒体信息的输入。根据学习过程中特征提取的不同,学习机制可以分为两类:笔画密度学习机制和灰度投影学习机制。并且针对在模板匹配过程中出现的阈值选择问题,采用学习机制实现对阈值的自动确定。
The technology to recognize Chinese character in image is a practical technology that can automatically recognize the Chinese character by using computers. It is widely used in different areas. The character recognition technology can also be applied to childhood education, which provides the young children a new human-computer interaction form .In additional, the multimedia technology can provide the children a good Chinese character learning environment full of images, text and sound. Thus the character recognition technology and multimedia technology can combine to form the Chinese characters recognition technology particularly for multi-media learning. It can not only identify the character image automatically, but also it can present the character images in various forms of multimedia information such as pictures, sounds, video, audio, animation and so on, which can provide children a good learning environment full of images, text and sound. It makes up the shortcomings like the emptiness and uncertainty of the shape and meaning of Chinese characters in traditional education, and also the shortcomings of rote learning. In addition, it can reduce the difficulty of Chinese characters learning, enhance the practical ability of students, and promote children to learn more actively, seriously and attentively, which in all provide a more effective way for children to learn Chinese characters.
Based on the research of the Chinese character image recognition technology, and combined with multimedia information, this paper realizes the representation of the character image recognition and multimedia information. Then, these technologies are used to develop a system prototype for Chinese character learning. And the algorithm in this paper proves to be effective after the system test.
The main contents are as follows:
(1) Feature selection and the design of classifier. Based on the features of Chinese characters, the density of stroke and the gray-scale projection are extracted to be the features of Chinese characters. And the method of feature extraction by dynamic classification is applied according to the different stroke scopes. Then, different classifiers are selectively designed according to the different scopes of feature extraction. Based on the feature of the gray-scale projection, this paper introduces the method of template matching of gray-scale projection to realize the Chinese character image recognition.
(2) The conflict management mechanism. The conflict management mechanism is used to solve the problem of Chinese characters conflicts that appear in the process of Chinese characters image recognition. The conflicts mainly can be divided into two cases: multi-codes for sole-character and multi-characters for sole-code and. According to different scenarios, this paper will adopt different conflict management mechanism.
(3)The learning mechanism. To solve the problem that some Chinese characters can't be identified in the process of Chinese character recognition, this paper uses the learning mechanism to extract the characteristics of Chinese characters and input the multimedia information. According to the difference of the extracted feature during the learning process, the learning mechanism can be divided into kinds that are the stroke-density learning mechanism and the gray-scale projection learning mechanism. Meanwhile, As to the problem of threshold choice in the process of template matching, this paper chooses the corresponding learning mechanism to determine the threshold automatically.
At last this paper develops the system prototype based on the above algorithm and the results prove that the system can effectively identify the image of Chinese characters, and can aid children's learning by multimedia information.
引文
[1]倪桂博.印刷体文字识别的研究.[硕士论文].华北电力大学计算机应用技术,2007:1-18.
[2]申小龙.论汉字的文化定义.汉字文化,2003.(2):8-15.
[3]李静.幼儿汉字多元化教育研究.[博士论文].西南师范大学教育学原理,2005:1-6.
[4]李静.汉字构形特征与幼儿认知的共鸣.儿童发展与教育,2006(7-8):45-48
[5]http://article.pchome.net/content-77559.html
[6]李定荣.多媒体计算机创设幼儿语言学习情境研究.[硕士论文].南京:南京师范大学教育技术学,2006:14.
[7]http://bbs.szhome.com/quote.aspx?id=38498081"e=38498081
[8]宋祥君,王春友,陈燕.基于三种不同技术的多媒体系统登录模块技术研究.计算机与现代化,2004,110(10):138-140.
[9]张忻中.汉字识别技术.北京:清华大学出版社,1992
[10]张德喜.手写体汉字机器识别技术的现状分析.许昌师专学报,1999,18(3):91-95.
[11]R.Casey,G.Nagy.Recognition of printed Chinese character.IEEE T.Elec Comp ut,1966,1(15).91-101.
[12]钱自拓.汉字图像识别研究.[硕士论文].合肥:合肥工业大学检测技术与自动化装置,2005:6.
[13]朱学芳,石青云,程敏德.文字识别中的图像自适应预处理.信号处理,1997,13(2):132-140.
[14]姜莎莎.文档图像识别系统的设计与实现.[硕士论文].武汉:华中科技大学,2004:12,14,5.
[15]刘炜.基于图像处理的车牌识别技术的研究.[硕士论文].哈尔滨:哈尔滨工业大学,2008:13.
[16]侯艳平.脱机手写体汉字识别研究.[硕士论文].扬州:扬州大学,2008:4,13,14.
[17]贾永红.数字图像处理.第一版.武汉:武汉大学出版社,2003:22.
[18]韦平安.车牌识别系统技术研究.[硕士论文].桂林:广西师范大学,2005:15-16.
[19]杨淑莹.图像模式识别—VC++技术实现.北京:清华大学,北京交通大学出版社,2006:138,19.
[20]Steinherz T,Intrator N,Rivlin E.Skew detection via principal components analysis.Proceeding of ICDAR' 99,1999:153-156.
[21]CHEN Ming,DING Xiao-qing.A Robust skew detection algorithm for grayscale document image.Proceedings of ICDAR' 99,1999:617-620.
[22]Okun O.Severe document skew detection.SPIE Conference on Mathematical Modeling and Estimation Techniques in Computer Vision,1998,3457:263-274.
[23]吴昊.汽车牌照自动识别系统.[硕士论文].电子科技大学信号与信息处理,2008:25.
[24]何蕾.数字图像复原技术研究.[硕士论文].合肥工业大学计算机软件与理论,2007:6.
[25]刘宁钟,杨静宇.基于投影算法的二维条码识别.计算机工程,2002,28(9):32-33.
[26]刘维平.中文印刷体文档内容识别系统研究.[硕士论文].哈尔滨工程大学模式识别与智能系统,2007:16,17.
[27]万励,陈洪波.表达式符号的大小归一化方法.广西大学梧州分校学报.2005,15(1):79-82.
[28]徐春明,张天平,王正群等.基于协同学的人脸分类集成.扬州大学学报:自然科学版,2006,9(2):48-52.
[29]边肇祺,张学工.模式识别.第2版.北京:清华大学出版社,2001.
[30]K.Z.Mao.Fast Orthogonal Forward Selection Algorithm for Feature Subset Selection.IEEE Trans.Neural Networks,2002,13(5):1218-1224.
[31]Hua-Liang Wei,Stephen A.Billings.Feature subset selection and ranking for data dimensionality reduction.IEEE Trans.Pattern Analysis and Machine Intelligence,2007,29(1):162-166.
[32]张昱.纸币号码图像识别方法研究.[硕士论文].沈阳工业大学,2005:44,45.
[33]陈伟.智能交通图像识别系统的研究.[硕士论文].浙江大学信息学院,2002:9.
[34]Cheng Lin Liu,In Jung Kim,Jin H Kim.Model-based stroke extraction and matching for handwritten Chinese character recognition.Pattern Recognition,2001(34):2339-2352.
[35]Y Mizukami.A handwritten Chinese character recognition system using Hierarchical displacement extraction based on directional features.Pattern Recognition Letters,1998,(19):595-604.
[36]冯志敏.基于结构特征的手写体汉字识别研究.[硕士论文].云南师范大学,2006:8.
[37]潘志杰,任赛君,陈友荣.基于BP神经网络的单数字字符识别算法实现.电脑知识与技术,2008,29(2):461-463.
[38]董二林.多媒体数据库系统在教学中的应用与研究.[硕士论文].东北师范大学教育技术学,2006:22.
[39]http://www.onegreen.net/Article/Database/SQLServer/SQLServer9/15663.html
[40]贾建忠.脱机印刷体维吾尔文字识别特征选择和分类器设计方法的研究.[硕士论文].苏州大学计算机应用技术,2008:31.
[41]周长发.精通VC++图像处理编程.第三版.北京:电子工业出版社,2007:1.
[42]刘韬,楼兴华.SQL Server 2000数据库系统开发实例导航.北京:人民邮电出版社,2004:270.
[43]曹红根,丁勇.数据库应用系统开发实例.第一版.北京:清华大学出版社,北京交通大学出版社,2008:30.
[44]李逸波等.多媒体数据库技术.北京:机械工业出版社,2004.
[45]徐云彪.数据库原理与技术.杭州:浙江大学出版社,2004.
[46]叶成林,徐福荫,任光杰.多媒体数据库及其教育应用.中国电化教育,2003,200(9):100-103.
[47]王大力.基于车牌识别技术的城市车辆稽查系统研究与应用.[硕士论文].沈阳工业大学计算机技术,2008
[48]李华刚,郭壮辉,习志奇,胡柯.智能监控设备图像识别系统研究.系统仿真技术,2007,3(3):169-173.
[49]刘显润.关于多媒体系统在教学中的应用及思考.教改前沿,1
[50]王路群,曹静.多媒体系统中多通道用户界面的研究.江汉大学学报(自然科学版),2004,32(3):45-47.
[2]申小龙.论汉字的文化定义.汉字文化,2003.(2):8-15.
[3]李静.幼儿汉字多元化教育研究.[博士论文].西南师范大学教育学原理,2005:1-6.
[4]李静.汉字构形特征与幼儿认知的共鸣.儿童发展与教育,2006(7-8):45-48
[5]http://article.pchome.net/content-77559.html
[6]李定荣.多媒体计算机创设幼儿语言学习情境研究.[硕士论文].南京:南京师范大学教育技术学,2006:14.
[7]http://bbs.szhome.com/quote.aspx?id=38498081"e=38498081
[8]宋祥君,王春友,陈燕.基于三种不同技术的多媒体系统登录模块技术研究.计算机与现代化,2004,110(10):138-140.
[9]张忻中.汉字识别技术.北京:清华大学出版社,1992
[10]张德喜.手写体汉字机器识别技术的现状分析.许昌师专学报,1999,18(3):91-95.
[11]R.Casey,G.Nagy.Recognition of printed Chinese character.IEEE T.Elec Comp ut,1966,1(15).91-101.
[12]钱自拓.汉字图像识别研究.[硕士论文].合肥:合肥工业大学检测技术与自动化装置,2005:6.
[13]朱学芳,石青云,程敏德.文字识别中的图像自适应预处理.信号处理,1997,13(2):132-140.
[14]姜莎莎.文档图像识别系统的设计与实现.[硕士论文].武汉:华中科技大学,2004:12,14,5.
[15]刘炜.基于图像处理的车牌识别技术的研究.[硕士论文].哈尔滨:哈尔滨工业大学,2008:13.
[16]侯艳平.脱机手写体汉字识别研究.[硕士论文].扬州:扬州大学,2008:4,13,14.
[17]贾永红.数字图像处理.第一版.武汉:武汉大学出版社,2003:22.
[18]韦平安.车牌识别系统技术研究.[硕士论文].桂林:广西师范大学,2005:15-16.
[19]杨淑莹.图像模式识别—VC++技术实现.北京:清华大学,北京交通大学出版社,2006:138,19.
[20]Steinherz T,Intrator N,Rivlin E.Skew detection via principal components analysis.Proceeding of ICDAR' 99,1999:153-156.
[21]CHEN Ming,DING Xiao-qing.A Robust skew detection algorithm for grayscale document image.Proceedings of ICDAR' 99,1999:617-620.
[22]Okun O.Severe document skew detection.SPIE Conference on Mathematical Modeling and Estimation Techniques in Computer Vision,1998,3457:263-274.
[23]吴昊.汽车牌照自动识别系统.[硕士论文].电子科技大学信号与信息处理,2008:25.
[24]何蕾.数字图像复原技术研究.[硕士论文].合肥工业大学计算机软件与理论,2007:6.
[25]刘宁钟,杨静宇.基于投影算法的二维条码识别.计算机工程,2002,28(9):32-33.
[26]刘维平.中文印刷体文档内容识别系统研究.[硕士论文].哈尔滨工程大学模式识别与智能系统,2007:16,17.
[27]万励,陈洪波.表达式符号的大小归一化方法.广西大学梧州分校学报.2005,15(1):79-82.
[28]徐春明,张天平,王正群等.基于协同学的人脸分类集成.扬州大学学报:自然科学版,2006,9(2):48-52.
[29]边肇祺,张学工.模式识别.第2版.北京:清华大学出版社,2001.
[30]K.Z.Mao.Fast Orthogonal Forward Selection Algorithm for Feature Subset Selection.IEEE Trans.Neural Networks,2002,13(5):1218-1224.
[31]Hua-Liang Wei,Stephen A.Billings.Feature subset selection and ranking for data dimensionality reduction.IEEE Trans.Pattern Analysis and Machine Intelligence,2007,29(1):162-166.
[32]张昱.纸币号码图像识别方法研究.[硕士论文].沈阳工业大学,2005:44,45.
[33]陈伟.智能交通图像识别系统的研究.[硕士论文].浙江大学信息学院,2002:9.
[34]Cheng Lin Liu,In Jung Kim,Jin H Kim.Model-based stroke extraction and matching for handwritten Chinese character recognition.Pattern Recognition,2001(34):2339-2352.
[35]Y Mizukami.A handwritten Chinese character recognition system using Hierarchical displacement extraction based on directional features.Pattern Recognition Letters,1998,(19):595-604.
[36]冯志敏.基于结构特征的手写体汉字识别研究.[硕士论文].云南师范大学,2006:8.
[37]潘志杰,任赛君,陈友荣.基于BP神经网络的单数字字符识别算法实现.电脑知识与技术,2008,29(2):461-463.
[38]董二林.多媒体数据库系统在教学中的应用与研究.[硕士论文].东北师范大学教育技术学,2006:22.
[39]http://www.onegreen.net/Article/Database/SQLServer/SQLServer9/15663.html
[40]贾建忠.脱机印刷体维吾尔文字识别特征选择和分类器设计方法的研究.[硕士论文].苏州大学计算机应用技术,2008:31.
[41]周长发.精通VC++图像处理编程.第三版.北京:电子工业出版社,2007:1.
[42]刘韬,楼兴华.SQL Server 2000数据库系统开发实例导航.北京:人民邮电出版社,2004:270.
[43]曹红根,丁勇.数据库应用系统开发实例.第一版.北京:清华大学出版社,北京交通大学出版社,2008:30.
[44]李逸波等.多媒体数据库技术.北京:机械工业出版社,2004.
[45]徐云彪.数据库原理与技术.杭州:浙江大学出版社,2004.
[46]叶成林,徐福荫,任光杰.多媒体数据库及其教育应用.中国电化教育,2003,200(9):100-103.
[47]王大力.基于车牌识别技术的城市车辆稽查系统研究与应用.[硕士论文].沈阳工业大学计算机技术,2008
[48]李华刚,郭壮辉,习志奇,胡柯.智能监控设备图像识别系统研究.系统仿真技术,2007,3(3):169-173.
[49]刘显润.关于多媒体系统在教学中的应用及思考.教改前沿,1
[50]王路群,曹静.多媒体系统中多通道用户界面的研究.江汉大学学报(自然科学版),2004,32(3):45-47.