手写数字特征的提取与分析

毕业设计(论文)题目手写数字特征提取与分析专业电子信息工程班级084班姓名梁杰指导教师周扬（讲师）所在学院信息学院完成时间：2012年5月承诺书我谨此郑重承诺：本毕业设计（论文）是本人在指导老师指导下独立撰写完成的。凡涉及他人观点和材料，均依据著作规范作了注释。如有抄袭或其它违反知识产权的情况，本人愿接受学校处分。承诺人（签名）：年月日手写数字特征提取与分析信息科技学院电子信息工程专业梁杰摘要：目前，模式识别领域在日常生活中的应用已经越来越广泛，比如人脸、指纹识别，字符识别，车牌识别。所以，对数字识别进行学习与研究是非常有必要的。本课题为数字字符识别模拟演示系统。主要是利用正态分布下的最小错误率Bayes方法和最小风险Bayes方法，来实现手写数字从0到9的识别。该系统首先是实现模拟手写数字；然后利用轮廓特征法将5*5的模板提取出样品的特征，采用模板可以使同一形状、不同大小的样品得到归一化的特征提取，所以有能力对同一形状、不同大小的样品视为同类；最后结合Bayes决策进行判别。使用最小错误率Bayes方法，在判别过程中能使错误率达到最小，即使错分类出现的可能性最小，而最小风险Bayes方法，在判别过程中可以使风险达到最小，减少危害大的错分类情况。本设计是利用Matlab实现的，实验证明，该系统对于模拟手写的数字基本上能正确识别，但是对于手写不规范的数字会存在错判的情况，这跟样品库的有限有关。关键词：模式识别；最小错误；最小风险；特征选择；模拟手写；Matlab实现HandwrittendigitalfeatureextractionandanalysisLiangJie，Electronicandinformationengineering,CollegeofInformationScienceandTechnologyAbstract:Atpresent,thefieldofpatternrecognitionineverydaylifehasbeenmoreandmorewidelyused,suchastheface,fingerprintrecognition,characterrecognition,vehiclelicenseplaterecognition.Therefore,thedigitalidentificationoflearningandresearchisverynecessary.Thetopicforthedigitalcharacterrecognitionsimulationdemosystem.MainlyusingnormaldistributionundertheminimumerrorrateofBayesmethodandBayesmethodtoachievetheminimumrisk,handwrittendigitsfrom0to9oftheidentification.Thesystemfirstistorealizethesimulationofhandwrittennumeral;thenusingcontourfeaturewillbe5*5templatesextractedsamplecharacteristics,usingthetemplateinthesameshape,differentsizesofsamplestobenormalizedfeatureextraction,sotheabilityofthesameshape,differentsizesofsamplesassimilar;finallycombinedwiththeBayesdecisiondiscriminant.MinimumerrorrateusingtheBayesmethod,thediscriminationprocesscanmaketheerrorratereachesaminimum,evenwrongclassificationandthepossibilityoftheminimum,whileminimizingrisksBayesmethodinjudgingprocess,canmaketheriskminimum,harmreductioninfaultclassification.ThisdesignistheuseofMatlabtoachieve,experimentsshowthat,thesystemforthesimulationofhandwrittendigitalbasicallycorrectidentification,butforhandwritingirregularnumbermayhavemisjudgedcase,thiswiththesamplelibraryassociation.Keywords:Patternrecognition；minimumerror；minimumrisk；featureselection；simulatedhandwriting；Matlab目录1绪论..................................................................................................................................11.1手写数字特征提取与分析的背景与意义..................................................................11.2手写数字特征的识别技术简介..................................................................................11.3现有的手写特征提取的有关算法..............................................................................21.4手写特征的典型应用..................................................................................................21.5本文研究的内容..........................................................................................................32模式识别与MATLAB的介绍..........................................................................................42.1模式识别.....................................................................................................................42.1.1模式识别的基本概念..........................................................................................42.1.2模式识别系统......................................................................................................42.1.3相关值计算..........................................................................................................42.2MATLAB.........................................................................................................................52.2.1Matlab软件的介绍................................................................................................52.2.2Matlab的主要优缺点...........................................................................................62.2.3Matlab图像类型及转换分析................................................................................73手写特征的提取与选择....................................................................................................93.1特征的种类与筛选......................................................................................................93.1.1笔划密度特征.......................................................................................................93.1.2傅立叶变换特征...................................................................................................93.1.3轮廓特征.............................................................................................................113.1.4投影特征.............................................................................................................123.1.5重心及重心矩特征.............................................................................................143.1.6首个黑点位置特征.............................................................................................143.1.7粗网格特征.........................................................................................................153.2特征提取方法............................................................................................................153.2.1结构特征提取方法.............................................