基于手繪草圖的工程圖檔檢索技術(shù)

1、南京航空航天大學(xué)碩士學(xué)位論文基于手繪草圖的工程圖檔檢索技術(shù)姓名：李雪峰申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：周良20081201基于手繪草圖的工程圖檔檢索技術(shù)iiAbstractWiththerapiddevelopmentoftheinfmationtechnologytheelectronicdrawingdocumentsaregraduallyplayingimptantrolesinallwalksofliferep

2、lacingthetraditionalpaperfiles.Howeveritisstilladifficultwktoobtainrequireddrawingdocumentsfromthedrawingdocumentsdatabasequicklyconveniently.Currentlythedrawingdocumentsretrievalisusuallyaccomplishedbasedonthekeywdssuch

3、asprojectnamedesignerdocumentnumber.Keywdbasedretrievaltechnologyisfastsimplewhereasmanuallabellingtakesalotoftroubles.Meoveritisnoteasytodescribethecontentindrawingdocumentswithwds.Asaresultinthispaperweproposeanovelske

4、tchbasedengineeringdrawingdocumentsretrievaltechniquewhichbelongstothecontentbasedimageretrievaltechnology.Itemphasizesonthegraphicfeaturesofdocumentsratherthankeywdstherefeitmakestheretrievalmedirectclear.Furthermeitism

5、econvenientfuserstoexpresstheirretrievalintentionwhenthehdrawnsketchesareuseddirectlyastheretrievalconditions.Afterdeepinvestigationonthesketchesrecognitiondrawingdocumentsretrievaltechniquethispaperpresentsasystemframew

6、koftheengineeringdrawingdocumentsretrievaltechnologybasedonhdrawnsketchesrealizesaprototypesystemaswell.Themainresearchesare(1)Proposesasketchrecognitionschemebasedonadaptivevertexdetection.WebringSupptVectMachineintothe

7、sketchrecognitionrealizetheuseradaptivevertexdetectionconsequentlytherecognitionrateisimproved.(2)Proposesthemultiplestagestructureoftheelectronicdrawings.Basedonthebasicelementstagestructureweaddthecombinationelementsta

8、gestructurewhichisindependentofthegraphic’sdrawingmethods.Thisameliationtosomeextentresolvestheproblemofmultiunderstingofoneimageimprovestherecallratioprecisionratioatthesametime.(3)Putsfwardamultifeaturefilteringretriev

9、alscheme.Weconsiderthreekindsoffeaturesi.e.wecombineshapetopologyaswellasspatialrelationshiptofilterthedatabaseimagesgraduallynarrowthesearchscopeconsequentlyweimprovetheefficiencyoftheretrieval.Keywds:sketchrecognitionS