三维模型实拍纹理高精度重建研究

《三维模型实拍纹理高精度重建研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、浙江大学计算机科学与技术学院博士学位论文三维模型实拍纹理高精度重建研究姓名：刁常宇申请学位级别：博士专业：计算机应用技术指导教师：鲁东明20071019堑至查竺鲤主兰堡堡塞!!曼通遥对三缭模型实籀纹理高鞲度重建技术的磷究，开发建立了离精度绞瑗重建系统，并在文物数字化存档的实际工作中进行了应用验证。从敦煌、金沙、良渚、澍姆渡等遗址多律珍贵文物三维模型实拍纹璞高精度踅建的结聚来看，本文的研究成果有效解决了基于实拍照片映射和融合生成三维模型高分辨率离保真纹理的问题。关键词：三维建模，纹理映射、重光照、纹理融合、BRDF光场模型新疆支学糖士学位论文AbstractRealobj

2、ects’3DreconstructionisoneofthemostimportantLssuesinVirtualRealityresearcharea．Itisbeingusedmoreandmorewidelyinmanyfields,suchasdigitalcollectionofculturalheritages，film，television，advertisement，game,ere．Theresearcholi3Dshapereconstructionandtexturereconstructionfromphotographaretwomaina

3、spectsofmodelingrealobjects．For3Dshapereconstruction，3Dscan，3Dmodelalignmentand3Dmodelmergingtechnologiescanachievehi曲precisionresultsflow．However,fortexturereconstructionfromphotograph，texturemappingandfusingtechnologies8襻stillunderdeveloping．Highprecisiontexturereconstructionfromphotog

4、raphwillfocusonrebuildingintegratedtexturefromseveralphotographsofrealobjectfora3Dmodel，andtrytopreservethetexlalre’sresolutionandcolorstabilitywhileprocessing．Inthethesis,theentiretextamreconstructionprocedureanditsrelatedkeytect-mofogiesareanalyzedandsummarized．And,ahighprecisiontextur

5、ereconstructionsolutionwillbepresentedindetail，whichmainlyconsistsoffourparts，asfollowing．1)ReviewoftexturereconstructiontechnologiesThedevelopmenthistoryoftexturereconstructiontechnologiesisreviewed，And，thecam∞,dcalibration，textureanti—aliasing，textarefusingandBRDFreconstructiontechnolo

6、giesareintroducedandcharacterized．Thenewestresearchresultsofthesetechnologiesarealsopresented．2)ArbitrarycorrespondingpointsbasedtexturemappingArbitrarycorrespondingpointsbasedtexturemappingincludesinteractivetexturemappingbasedollarbitrarycorrespondingpointsbetweenthesurfaceof3Dmodeland

7、texturephotograph，fastself-occlusioncullingbased0ntheextrinsiccameraparameters，andtexturecoordinatesrectificationbasedoncorrespondingpoints．3)Globaltextureretigtltingand趣Globaltexturerelightingandfusingincludesglobalrelightingofmultipletexturesbasedonleastmean—squareerror