基于标注信息学习的植物多样化建模分析-modeling and analysis of plant diversity based on annotation information learning.docxVIP

ResearchonAnnotation-learning-basedDiversified TreeModelingABSTRACTVirtualrealitytechnologyhasalwaysbeenahotspotincomputergraphics.As anefficientwaytorepresentobjectsinvirtualworld,arapidreconstructionof three-dimensional(3D)modelisgainingextensiveattention.Beingacommon phenomenon,the diversityoftreesdeterminesthecomplexity of their corresponding3Dmodels,andtherealismoftreemodelshasanstronginfluenceon theimmersionofvirtualview.Asaresult,thesimulationoftreegrowthandthe modelingoflandscapehavebeenthefocusofstudy.Thoughautomaticmodeling techniqueachievesquitegoodeffect,usercanonly define effective generation ruleswithspecializedknowledgeoftrees,anditisdifficultforusertounderstand theoverallstructurebecauseoftheabstractmodelingprocess.Interaction-based methodcancontrolthetreemodelingprocessconveniently,buttheresultisgreatly affectedbyhumanexperience,anduserhastorepeattheprocessesforeachtree. Therefore,howtomodelmasstreeswithdifferentstructuralfeaturesbyusinga unifiedtechniqueisanimportant problem.Basedoninverseproceduralmodeling,therapidautomatic3Dtreemodeling methodhasbeenstudiedinthisdissertationbyintroducinglearningmechanismand takinguser-identifiedinformationasguidance,anditmainlyincludesthefollowing points:Keyinformationwhichinfluencestreeappearanceisdefinedbasedonan investigationofbranchesinbotany; annotations are added to the tree structureaccordingtothesekeyinformation;aBayesNetworkfeaturedatabase oftreesisestablishedbyextractingfeaturesfromaplentyofpictures.ProductionrulesofL-systemsaredefinedaccordingtothebranchcharactersof treesinbotanyafterstudyingmodelingmethodsbasedonL-systems;theBayes Networkisusedtoconductthemodelingprocessundertheguideoffeature database.Leavesmodelingofdifferenttreesarerealizedbyusingthetransparenttexture andleaveslayoutalgorithmafterexploringthemodelingtechniqueforsingle leaf;thismodelingmethodisintegratedtotheannotation-learning-based diversifiedtreemodelingsystem.Keyword: TreeModeling;L-systems;BayesNetwork;Annotation;Inverse Proc