简介:Inthispaper,wepresentanovelSupportVectorMachineactivelearningalgorithmforeffective3Dmodelretrievalusingtheconceptofrelevancefeedback.Theproposedmethodlearnsfromthemostinformativeobjectswhicharemarkedbytheuser,andthencreatesaboundaryseparatingtherelevantmodelsfromirrelevantones.Whatitneedsisonlyasmallnumberof3Dmodelslabelledbytheuser.Itcangrasptheuser'ssemanticknowledgerapidlyandaccurately.Experimentalresultsshowedthattheproposedalgorithmsignificantlyimprovestheretrievaleffectiveness.Comparedwithfourstate-of-the-artqueryrefinementschemesfor3Dmodelretrieval,itprovidessuperiorretrievalperformanceafternomorethantworoundsofrelevance
简介:现在的纸为与快速的3D片和格子生成器(RAPID3DGRID)相结合的turbomachinery的空气动力学的设计描述优化方法论,一个N.S解答者,一个片parameterization方法(BPM),一个基于坡度的analyseparameterization方法(GPAM),有激增的一个反应表面方法(RSM)算法和一个简单坡度方法。由片parameterization方法的使用,一个接近音速的压缩机转子能被一套多项式表示,然后它使我们能转变表示坐标的片数据到表示参数然后减少参数的数字。与改变任何一个参数,我们能获得几组样品。这里,仅仅十个参数被认为寻找一个优化压缩机转子。由于优化,断热的效率被1.73%增加。
简介:Theshiftinthepercolationthresholdofcompressedcompositeswasstudiedbya3Dcontinuumpercolationmodel.AMonteCarlo(MC)methodwasemployedinthesimulations.Thepercolationthresholdwasfoundtorisewiththecompressionstrain,whichcapturesthebasictrendincompression-inducedconductivityvariationfromtheexperiments.Bothfiberbendingandtextureformationcontributetothepercolationthreshold.Theresultssuggestthatfillerswithahighaspectratioaremoredesirableforsensorandelectricalswitchapplications.
简介:
简介:Inthiswork,anewmethodtodealwiththeunconnectedpixelsinmotioncompensatedtemporalfiltering(MCTF)ispresented,whichisdesignedtoimprovetheperformanceof3Dliftedwaveletcoding.Furthermore,multipledescriptionscalablecoding(MDSC)isinvestigated,andnovelMDSCschemesbasedon3Dwaveletcodingareproposed,usingtheliftingimplementationoftemporalfiltering.TheproposedMDSCschemescanavoidthemismatchprobleminmultipledescriptionvideocoding,andhavehighscalabilityandrobustnessofvideotransmission.Experimentalresultsshowedthattheproposedschemesarefeasibleandadequatelyeffective.
简介:Anapproachforgeneratinginteractive3Dgraphicalvisualizationofthegeneticarchitecturesofcomplextraitsinmultipleenvironmentsisdescribed.3Dgraphicalvisualizationisutilizedformakingimprovementsontraditionalplotsinquan-titativetraitlocus(QTL)mappinganalysis.Interactive3DgraphicalvisualizationforabstractexpressionofQTL,epistasisandtheirenvironmentalinteractionsforexperimentalpopulationswasdevelopedinframeworkofuser-friendlysoftwareQTLNetwork(http://ibi.zju.edu.cn/software/qtlnetwork).Noveldefinitionofgraphicalmetasystemandcomputationofvirtualcoordinatesareusedtoachieveexplicitbutmeaningfulvisualization.Interactive3DgraphicalvisualizationforQTLanalysisprovidesgeneticistsandbreedersapowerfulandeasy-to-usetooltoanalyzeandpublishtheirresearchresults.
简介:Thispaperisconcernedwithexperimentalandnumericalresearchon3DflowpastprismaticturbinecascadeSE1050(knowninQNETnetworkasopentestcaseSE1050).Theprimarygoalwastoassesstheinfluenceoftheinletvelocityprofileontheflowstructuresintheinterbladechannelandontheflowfieldparametersatthecascadeexitandtocomparethesefindingstoresultsofnumericalsimulations.Investigationsof3Dflowpastthecascadewithnon-uniforminletvelocityprofilewerecarriedoutbothexperimentallyandnumericallyatsubsonic(M2is=0.8)andattransonic(M2is=1.2)regimeatdesignangleofincidence.Experimentaldatawasobtainedusingatraversingdevicewithafive-holeconicalprobe.Numerically,the3DflowwassimulatedbyopensourcecodeOpenFOAMandin-housecode.AnalysesofexperimentaldataandCFDsimulationshaverevealedthedevelopmentofdistinctivevortexstructuresresultingfromnon-uniforminletvelocityprofile.Originofthesestructuresresultsinincreasedlossofkineticenergyandspanwiseshiftofkineticenergylosscoefficientdistribution.Differencesfoundbetweenthesubsonicandthetransoniccaseconfirmearlierfindingsavailableintheliterature.ResultsofCFDandexperimentsagreereasonablywell.
简介:在这份报纸,我们在场一三维(3D)在为收音机频率的晶片水平的真空包装技术微机电的系统(RFMEMS)共鸣器,低损失的硅vias在被用来播送RF发信号。Au-Sn焊接结合被采用象电的导体一样提供真空封装。封装帽子的一个RF模型被建立评估包装的寄生效果,它提供3DRFMEMS封装的一个有效设计答案。与建议包装结构,24dB的signal-to-background比率(SBR)被完成,以及共鸣器的优秀因素(Q因素)在包装以后从8000~10400增加。包装共鸣器有线性频率温度(英尺)在在0楨?慰数?之间的一个温度范围特征?牵敶?牡?牰癯摩吗??
简介:Inthispaper,weproposeahighlyautomaticapproachfor3Dphotorealisticfacereconstructionfromasinglefrontalimage.Thekeypointofourworkistheimplementationofadaptivemanifoldlearningapproach.Beforehand,anactiveappearancemodel(AAM)istrainedforautomaticfeatureextractionandadaptivelocallylinearembedding(ALLE)algorithmisutilizedtoreducethedimensionalityofthe3Ddatabase.Then,givenaninputfrontalfaceimage,thecorrespondingweightsbetween3DsamplesandtheimagearesynthesizedadaptivelyaccordingtotheAAMselectedfacialfeatures.Finally,geometryreconstructionisachievedbylinearweightedcombinationofadaptivelyselectedsamples.Radialbasisfunction(RBF)isadoptedtomapfacialtexturefromthefrontalimagetothereconstructedfacegeometry.Thetextureofinvisibleregionsbetweenthefaceandtheearsisinterpolatedbysamplingfromthefrontalimage.Thisapproachhasseveraladvantages:(1)Onlyasinglefrontalfaceimageisneededforhighlyautomaticfacereconstruction;(2)Comparedwithformerworks,ourreconstructionapproachprovideshigheraccuracy;(3)ConstraintbasedRBFtexturemappingprovidesnaturalappearanceforreconstructedface.