毕业设计基于人工神经网络的手写识别系统的设计与实现

基于人工神经网络的手写识别系统I本科毕业论文（20届）基于人工神经网络的手写识别系统的设计与实现所在学院专业班级计算机科学与技术学生姓名指导教师完成日期II摘要信息技术的快速发展，计算机迅速走进人们的生活，手写识别技术和应用领域越来越广泛，比如在移动设备上、文印工作中都离不开它的身影，极大的方便了人们的日程生活、工作、学习，因此手写识别具有很强的应用价值和实用价值。本文中利用神经网络算法识别数字图像和手写汉字。识别的前提是获得稳定的神经网路，需要通过大量的训练样本进行训练。针对数字图像首先对图像进行处理，通过灰度化、二值化、中值滤波、梯度锐化、归一化等算法获得特征矩阵，然后用经过训练的神经网路算法识别需要识别的特侦矩阵，进而获得识别的结果。针对汉字识别，首先处理图像获得特征矩阵，通过经过训练的神经网络的识别获得笔画，然后用笔画去识别汉字，最后获得识别的汉字以及联想汉字和词组。经过测试和验证，基于人工神经网络的手写识别系统能够有效识别数字手写图像与手写汉字，在本文选用的一些测试文字与数字图像的识别上获得了不错的结果。关键词：手写识别；图像处理；神经网络第1章绪论IIIABSTRACTTherapiddevelopmentofinformationtechnologymakesthecomputerimmediatelyappearinpeople'slives.Theapplicationofhandwritingrecognitiontechnologybecomeswidespread,andtherearemoreandmoreapplicationareas.Forexample,wecanseeitonamobiledeviceandintheprintingwork.Itmakespeople'sdailylife,workandlearningmoreconvenient.Therefore,thehandwritingrecognitionhasverystrongappliedvalueandpracticalvalue.Inthisarticle,thewriterusesneuralnetworktoidentifydigitalimageandhandwritingChineseideogram.Thepremiseofrecognitionistoobtainastableneuralnetworkandagreatdealofsampletrainingisneeded.Fordigitalimage,weneedtoprocesstheimages,andgetcharacteristicmatrixthroughaseriesofalgorithms,includinggraying,linearization,Medianfilter,gradientsharpeningandnormalization.ForChineseideogramrecognition,weshouldgainacharacteristicmatrixbyprocessingimages,andthengetpaintingpenthroughthetrainingneuralnetworkidentification.Next,weusepaintingpentorecognizeChineseideogram.Atlast,theChineseideogramandassociativeChinesecharactersandphraseswillbeobtained.Aftertestingandverification,handwritingrecognitionwhichisbasedonartificialneuralnetworksystemcaneffectivelyidentifythehandwritingimageandChineseideogram.Weobtainsatisfyingresultsintheidentificationtestonwordsanddigitalimage.Keywords:Handwritingrecognition;imageprocessing;neuralnetwork基于人工神经网络的手写识别系统IV目录第1章绪论...............................................................................................................................11.1课题研究的背景..........................................................................................................11.2课题研究的目的及意义..............................................................................................11.3国内外研究现状..........................................................................................................21.4课题研究内容..............................................................................................................21.5论文的组织结构..........................................................................................................3第2章神经网络算法的原理...................................................................................................42.1神经网络的原理..........................................................................................................42.1.1生物神经元网络...............................................................................................42.1.2人工神经元网络...............................................................................................42.2神经元学习算法..........................................................................................................62.2.1前馈神经网络...................................................................................................62.2.2感知机...............................................................................................................62.2.3反向传播算法...................................................................................................82.2.4神经网络在模式识别上面的优势..................................................................112.3本章小结.....................................................................................................................11第3章系统设计与实现.........................................................................................................123.1神经网络算法的实现................................................................................................123.1.1神经网络的结构.............................................................................................123.1.2算法的结构.....................................................................................................123.2神经网络识别手写数字............................................................................................143.2.1手写数字的常规预处理方法.........................................................................143.2.2图像的灰度化.................................................................................................153.2.3图像的二值化.................................................................................................153.2.4中值滤波.........................................................................................................163.2.5去掉离散点.....................................................................................................163.2.6图像锐化.........................................................................................................173.2.7数字的切分.....................................................................................................18第1章绪论V3.2.8数字的归一化........