三維模型特征提取和相關(guān)反饋算法研究與實(shí)現(xiàn).pdf

1、西北大學(xué)碩士學(xué)位論文三維模型特征提取和相關(guān)反饋算法研究與實(shí)現(xiàn)姓名：張飛申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)軟件與理論指導(dǎo)教師：周明全20100621ResearchandImplementationofFeatureExtractionandRelevantFeedbackof3DModelABSTRACTTherapiddevelopmentofthethreedimensional(3D)modelingtechnologyandInte

2、mettechnologieshasledtoanincreasingamountof3Dmodels，bothontheIntemetandindomainspecificdatabaseAnd3Dmodelsplayaveryimportantroleinmanyfields，such嬲molecularbiologyculturalheritageprotectioncomputeraideddesignandmanufactur

3、ingandSOonThus，thispaperarriveaccuracyandefficientlyretrieveforthemassof3Dmodels，throughresearchingthetechnologyoffeatureextractionandrelevancefeedbackThemainworkandprogressarelistedasfollows：1)ImplementtheimprovedPCAand

4、isotropicpreprocessingalgorithmandwealsocompareandanalyzetheexperimentalresultsTheexperimentalresultsshowthatpreprocessing3Dmodelsbeforethefeatureextractioneffectivelyimprovestheaccuracyofthesearchalgorithm2)Improveanext

5、ractionalgorithmbasedonTerminalProjectionTransformationFromtheexperimentalresults，comparewithasimpleextractionalgorithmofthepixeldistributehistogrambasedonthreeviewsprojectionthealgorithmhasagoodabilitytodistinguishdiffe

6、rent3Dmodels，andeffectivelyimprovestheaccuracyof3Dmodelretrieval3)Proposeanewfeatureextractionalgorithmfor3DmodelbasedonRangeImageFirstthisalgorithmcomputes3Dmodelsprojectionrangeimagesonsixdirections，andthencomputesedge

7、directionhistogramandZemikemomentoftheserangeimagesAtlastcombiningthetwofeaturematches3DmodelsTheexperimentsshowthatthealgorithmavoidsthetraditionalvisualshortcomingsofthelossofspatialinformationandimprovestheaccuracyoft

8、heretrievalandrobustnesseffcctively4)ThispaperintroducestheweightSVMtelevancefeedbackbecauseofthesemanticgapbetweenthehighlevelsemanticsandlowlevelfeaturesTheexperimentalresultsindicatethatthealgorithmeliminatesthesemant