基于支持向量機(jī)與集成學(xué)習(xí)的紋理合成及分類.pdf

1、中國海洋大學(xué)碩士學(xué)位論文基于支持向量機(jī)與集成學(xué)習(xí)的紋理合成及分類姓名：王冉申請學(xué)位級別：碩士專業(yè)：計算機(jī)軟件與理論指導(dǎo)教師：董軍宇201106TexturesynthesisandclassificationBasedonsupportvectormachinesandensemblelearningAbstractTexturesynthesisandclassificationareactiveresearchto霉i岱during

2、thepaStfewyears，theyarewidelyusedinimageprocessing，computervisionandpatt鋤recognitionfieldTheaimoftexturesynthesisistoproduceanewimagewhichislargerthallnles鋤pleimage，andthenewonehasthesametexturecharacteristicswiththes鋤pl

3、eimageCurrenttechniquesof2Dtexturesynthesisarenearlyma嘶Unlike2DteXture，real‘world3Dsurfacetexturesnormallycompriseroughgeometrysu】血ceandvariousreflectancepropertiesThesamesurfacetextureCallvaryd珈naticallvwithdifferentill

4、uminationdirectionsandviewconditionsHowever2Dtextl鵬synthesismethodsignoretheeffectofilluminationandviewconditionsSotheresearchon3DsurfacetexturesynthesisismorerealisticAsaveryimportantb瑚dlofpatternrecognitionfieldtheaimo

5、ftextureclassificationistodistinguishthedifferenceofallobjoc岱withdifferenttexturecharact麗stics’Thisthesisintroducesthreestepsincludingtexturesynthesis，textureclaSsificationandtexturefeatureselectionInthisthesis，severalme

6、thodsabout2Dtexturesynthesis，featureextractionfeatureselectionandclassificationarefirstintroduced，andtlleIl廿1eresearchontexturesynthesisandclassificationisdiscussedUnliketexturesynthesisusin2supportvectormachines，whichca

7、nonlysynthesizehighlystructuredtextI鵬廿1e1)roposedmethodbasedonsupportvectormachinesandmultipleseedblockscaJlsuccessfullysynthesizebothrandomandstructuredtextures，anditis脅hereXtendedto3DsurfacetexturesynthesisBecausetrain