小麦腥黑穗病鉴定的svm方法本科论文.doc

本科毕业论文小麦腥黑穗病鉴定的SVM方法摘要纹理图像的自动分类在许多领域都是一项关键的任务，其中包括农作物产品等级分类、可视场景的目标检测、信息检索、医学应用等等。当直接在图像上进行操作时，传统的分类方法由于数据的高维特性表现差，很难取得较好的效果。但是支持向量机（SupportVectorMachine，SVM）可以克服极高维表示的缺陷，被广泛运用到纹理图像分类中去。本文所做的主要工作如下：1.简要分析支持向量机的工作原理，分析支持向量机核函数中各个参数对分类模型的影响，比较各参数对SVM的寻优能力。同时简单介绍SVM在纹理图像分类中的应用。2.对灰度图像使用灰度共生矩阵法、Tamura方法和Gabor滤波方法分别提取图像的纹理特征，并对比三种方法对分类出石纹和树皮纹理两种图像的能力。本文着重分析了灰度共生矩阵这一种方法。3.应用SVM对纹理图像分类的算法对矮腥和网腥纹理这两种纹理图像进行样本训练和分类，并对SVM的分类能力进行测试和比较，分析了不同参数优化方法对图像分类准确率的影响。关键词：支持向量机纹理图像特征图像分类腥黑穗病鉴定WheatredinbrandidentificationmethodofSVMLiZongshang(CollegeofEngineering,SouthChinaAgriculturalUniversity,Guangzhou510642,China)Abstract:Automaticclassificationoftextureimageisakeytaskinmanyfields,includingagriculturalproductsclassification,visualscenedetection,informationretrieval,medicalapplications,andsoon.Whenoperatingdirectlyontheimage,thetraditionalclassificationmethodasthedataofhighdimensioncharacteristicperformanceispoor,itisdifficulttoobtaingoodeffect.ButtheSupportVectorMachine(SVM)canovercomethedefectsveryhigh-dimensionalsaid,waswidelyusedinthetextureimageclassification.Inthispaper,themainworkdoneisasfollows:1.Thebriefanalysisoftheworkingprincipleofsupportvectormachine,analyzingeachparameterofkernelfunctionofsupportvectormachineontheclassificationmodel,theinfluenceofcomparingtheparametersoftheSVMoptimizationability.AtthesametimesimpleintroducestheapplicationofSVMinthetextureimageclassification.2.Forgrayimageusinggraylevelco-occurrencematrixmethod,TamuramethodandGaborfilteringmethodrespectivelytoextractimagetexturefeature,andcomparedthreemethodsofclassificationthestoneandbarktwotextureimage.Thispaperanalyzesthegraylevelco-occurrencematrixasakindofmethod.3.ApplicationofSVMfortextureimageclassificationalgorithmforDwarfbuntandStinkingsampletrainingandthesetwokindsoftextureimageclassification,andtheclassificationabilityofSVMfortestingandcomparison,analyzesthedifferentinfluenceontheaccuracyofimageclassificationparametersoptimizationmethod.Keywords:SupportvectormachineTextureimagefeaturesImageclas