1、UniversityCode：10225RegisterCode：080229DissertationfortheDegreeofMasterResearchonPatternRecognitionofWoodSurfaceColorCandidate：YangShaochunSupervisor：ProfWangKeqiAssociateSupervisor：AcademicDegreeAppliedfor：MasterofEngin

2、eeringSpeciality：AgriculturalElectrificationandAutomationDateofOralExamination：June，2008University：NortheastForestryUniversityAbstractAbstractColorisoneoftheimportantnaturalcharactersofwood，whichaffectsthevisualperceptio

3、nandevaluationofwoodandwoodproductsInviewofthattherewasnonationalstandardorindustrystandardinwoodindustrynowtodefineanddescribethewoodsurfacecolorimageprocessingtechnologyandpaRemrecognitionwasusedtostudyonthemethodtoimp

4、lementautomaticclassificationofwoodsurfacecolorAsetofparameterstodescribethewoodsurfacecolorandthemethodofpatternrecognitiontoclassifythewoodbasedontheseparameterswereestablishedItwouldprovideanimportantreferenceandtheor

5、eticalbasisforclassifyingwoodproductsanddecorationwoodbywoodslyfacecolorandprovidedasolidbasisoftheoryforwoodcolorclassificationautomaticallyCombiningwitllourresearchsubjectthewoodsampledatabasesusedforwoodsurfacecoloran

6、alyzingandclassifyingwerefoundedatfirstWithimageprocessingtechnologypreprocessingofthewoodimagesinsampledatabaseswasmadetoeliminatetheinfluenceofnoiseThreemethodofwoodsurfacecolorexpressionwasstudiedandthreekindsofcolorf

7、eaturewereextractedcorrespondinglyThemethodsoffeatureselectionandclassificationinthepatterrecognitionsystemwerealsostudied，andclassificationresultsofthreedifferentclassfiereswiththefeaturesbeforeandafterfeatureselectionw

8、asdiscussedWithcomprehensiveconsiderationtofeaturedimensionandclassificationresults，asetofwoodcolorparametersandthemethodofpattemrecognitionwereestablishedfinallyInviewofthecomplexi夠ofwoodcoloranditsnarrowdistributioninc

9、olorspace，threekindsofcolorfeaturewereproposedaftertheanalysisofwoodcolorexpression：①AsHSVcolorspacecouldreflecttheperceptionandidentificationofhumanbeingtocolorhistogramstatisticalfeaturesofthreeindependencevariablesHue

10、，SaturationandValuewereproposedtoexpresstheinformationofwoodcolor；②ColormomentsofHue，SaturationandValuewereusedtoexpresstheinformationofwoodcolor；③WiththecharacteristicsofhighresolvingpowerofL“abuniformcolorspace，fivecol

11、orfeatures，sevencolordifferencefeaturesandsevenstandarddeviationfeaturesofcolordifferencewerealsoproposedProblemsandmethodoffeatureselectioninpattemrecognitionsystemwerestudied，andclassificationaccuracyofNearestNeighbor(

12、NN)classiferwasadoptedasevaluationcriterionThetheoriesofSimulatedAnnealingAlgorithm(SA)andGeneticAlgorithm(GA)werestatedFeatureselectionsbasedonSAclassificationaccuracyandGAclassificationaccuracywererealized，andthenthese