金融票据OCR系统票据和字符的切分与识别-模式识别与智能系统专业论文.docxVIP

研究生毕业论文：票据和字符的切分与识别 研究生毕业论文：票据和字符的切分与识别 作者：于武贵 摘要 该论文是对我研究生期间对金融票据OCR系统及部分关键技术研究的总结。 分别在手写体与印刷体字体判断的特征提取与特征选择，利用遗传算法对相似字 进行特征选择，彩色票据的大写金额切分，基于统计方法的混排文字切分与分类 等方向研究上取得了一定的成果。第一章概述，主要讨论了票据OCR的背景和现 状，对票据OCR的一些关键技术作了简单介绍，以及本论文的主要内容；第二章 是对手写体与印刷体字体判断的特征提取与特征选择研究的总结，首先分析了手 写体与Fp届,J体区分的必要性，然后重点介绍了由我完成的特征选择部分，并对结 果进行了讨论；第三章是利用遗传算法对相似字进行特征选择的研究部分，先介 绍了特征选择和遗传算法的原理，然后分析了用遗传算法优选特征的方法，并讨 论了将所得到的优选特征应用于相似字的区分上的结果：第四章是彩色票据的大 写金额切分，介绍了彩色图像原理及切分主要手段以及用HSB方式的彩色票据 大写金额切分；第五章是基于统计方法的混排文字切分与分类的研究，介绍了简 单贝叶斯分类器的原理和分类器在切分中的应用，重点讨论了我完成特征提取部 分，最后对结果进行了分析。 [关键字] 票据0cR{模式识别；特征提取；特征选择：遗传算法：贝叶斯分类器：相 似字； 共48页 第1页 研究生毕业论文：票据和字符的切分与识别 研究生毕业论文：票据和字符的切分与识别 作者：于武贵 ABsTRACT ThiS themS iS a summary of my search in Banking Check OCR system and its key technologies during my two years study towards my Master’S degree．We applled genetic algorithms to feature extraction and selection in order to distinguish handwriting characters and printed characters． Other research was conducted in the following field：feature selectiOn of similar characters using genetic algorithm，segmenting of capital sum characters in colored check，segmenting and classification of mixed characters．We have made some achievement in our research and have several papers published．Chapter 1，overview，briefly introduces the background and current level of banking check OCR，with brief introduction to some key technologies．Chapter 2 summarized of feature extraction and selection in distinguishing of handwriting and printed characters．First the importance of distinguishing handwriting and printed characters was presented，and secondly some research in feature selection was introduced Chapter 3 brings forward the research of feature selection using genetic algorithm．The principle of feature selection and genetic algorithm was introduced，and the method of using genetic algorithm to select features was Dresented．In the end，the effect of using optimized features in the di stinguishing of similar characters was discussed．Chapter 4 in