复杂场景中的实时目标识别

上海交通大学硕士学位论文复杂场景中的实时目标识别姓名：孙伟斌申请学位级别：硕士专业：计算机软件与理论指导教师：肖双九;杨旭波20090101žþ°όÆa¬Æ Ø©EEE,,,|||µµµ¥¥¥¢¢¢žžž888III£££OOOÆÆÆ:þ°όÆXXX:^‡Æ???:B0603791ÆÆÆÒÒÒ:1060379012ïïïÄÄÄ))):š•RÆÆƉ‰‰;;;’’’:OŽÅ^‡†nØ“““:VÊRÅþþþ°°°ÏÏόŒŒÆÆÆ^^^‡‡‡ÆÆÆ2009ccc1ADissertationSubmittedtoShanghaiJiaoTongUniversityfortheMasterDegreeREALTIMEOBJECTRECOGNITIONINCOMPLEXENVIRONMENTAuthor:WeibinSunSpecialty:ComputerScienceAdvisor:Prof.ShuangjiuXiaoProf.XuboYangSchoolofSoftwareShanghaiJiaoTongUniversityShanghai,P.R.ChinaJanuary1,2009þ°όÆÆ Ø©M5(²x­(²µ¤¥Æ Ø©§´3“e§Õá?1ïÄ󊤤JØ©¥®²5²Ú^SN §ة؝¹?ÛÙ¦‡½8N®²uL½LŠ¬¤Jé©ïĉѭ‡z‡Ú8N§þ®3©¥±²(ªI²©¿£(²{Æ(Jd«úÆ Ø©Šö\¶µFϵcFþ°όÆÆ Ø©‡¦^ÇÖÆ Ø©Šö)þ°όÆk'3!¦^Æ Ø©5½§Ó¿Æ3¿•I[k'܀½ÅxØ©E‡Úf‡§#NØ©Ú/Çþ°όƌ±òÆ Ø©ܽܩSN?\k'êâ¥?1u¢§Œ±æ^K! ½×£E›ÃãÚ®?Æ Ø©£££———ØØØ©©©333)))———AAA„„„ÅÅÅddd555½½½¤¤¤Æ Ø©Šö\¶µ“\¶µFϵcFFϵcFE,|µ¥¢ž8I£OÁ‡EEE,,,|||µµµ¥¥¥¢¢¢žžž888III£££OOOÁÁÁ‡‡‡8I£O£ObjectRecognition¤ŠOŽÅÀú¥pÀú˜‡­‡ïįK§´óœU+¥¦OŽÅUPkaU嘫Ì‡{8I£O¦œUXÚUaú¿@£±Œ‚¸§éÅìXÚ!OŽÅ9Ϛ!œUzÅp±9¤kÙ§I‡OŽÅäkaUåXÚÑ´–'­‡UåCAc5§ŠOŽÅÀú˜‡ïÄ9:§8I£O3{†.¡Ñk­‡uЧ¿ÅìdïÄm©r•ó’A^©±ÄuÝKOry¢ïđ8µ§XúuŒ±A^uOry¢XÚ¥8I£OEâOry¢£AugmentedReality¤´˜«Œ±òJ[ÔN(Üy¢|µ¥Eâ§U|±^r†Ù?1p§l‰^r±Ù¸Nþ­Vʛc“±5§Ory¢®²¤J[y¢ïʇ­‡ïÄ+§ÓžéuÅp´˜‡#uА•éOry¢XÚI‡¢žpA:§©é¯„8IÆS±9¯„8I£OnØïÄ?1o(†&¢©Äké8I£O±9ƒ'+cïĆuÐ?1o(†äN`ã¶,é‘8µ§©ÛOry¢XÚ¥A^8I£OEâI¦†¯K§éäNI¦Ú¯K§(½©ïÄ̇8Iڇ¦µ¯„ÆS!¯„£¢ž¤£O!F~‚¸£E,|µ¤±9f©a£Œam§aS¤¶‘3é8I£O.†Eâ&?¥§âþãïÄ8I†‡¦Jј«M#(Ü8Iuÿ†f©a8I£O.SOAW£SmallObjectAsWords,¹Â´¦^.8IŠA.¤§¿3d.µee§Äu#8I£Oïħ)3‘8ïÄ¥é8IAJ†£ã˜‘M#¤J?1¯„8I£Oïħªˆ3F~‚¸¥¢ž8I£O¢(JL²§3©JÑ.µee§Äu©¥JÑ#.8IA†£ã§±9ƒA·ÜSOAW.¯„ÆSŽ{§©¤ïü@¢XÚÑUˆe¯„ÆS!¢žp8I£O'…cµ8I£O§8Iuÿ§OŽÅÀú§Ory¢§Auÿ§A£ã§©a-i-E,|µ¥¢ž8I£O=©Á‡RealTimeObjectRecognitioninComplexEnvironmentABSTRACTObjectrecognitionasahigh-levelvisionincomputervisionisanimportantresearchtopic.Itisoneofthemostimportantfieldsofartificialintelligencetogivethecomputertheabilitytohaveaperception.Objectrecognitionsystemcanmakeintelligentsystembeawarenessofsurroundingenvironment,whichisessentialtotherobotsystem,Computer-aidedmedicalcare,intelligenthuman-computerinteraction,aswellasallothersystemsthatneedscomputerstohaveaperceptionability.Inrecentyears,asahotspotincomputervision,objectrecognitionhasbeendevelopedsignificantlyinbothmethodandmodel,andgraduallybeginstobeappliedbyindustrialapplications.Basedontheprojectionbasedonaugmentedrealityresearchprojects,thispaperfocusesonobjectrecognitionresearchinAugmentedRealitySystem.AugmentedRealityisatechnologythatcancombinevirtualobjectswiththerealscenes,whichcanalsosupporthuman-computerinteractionsothatuserscanenjoyanimmersiveexperience.Sincethe1990s,AugmentedRealityhasbecomeanimportantareainvirtualreality,whileitisalsoanresearchtopicinHuman-ComputerInteraction.Aimingatthereal-timeinteractioninAugmentedRealitySystem,thispapersummarizesandexploresthefastmachinelearningmethods,aswellasrapidobjectrecognitionresearch.Firstly,thispapergivesafullsummarizationofthestate-of-artresearchanddevel-opmentinobjectrecognitionandrelevantfields.Thentherequirementsandproblemsinobjectrecognitioninaugmentedrealitysystemareanalyzedbasedonthebackgroundproject.Themaingoalandrequirementsofthispaperaredecidedbyanalyzingtheserequirementsandchallenges.Thispaperaimsat:FastLearning,Fast(Real-time)Recog-nitionincomplexenvironmentsandweakclassificationbecauseoflargedifferencebetweenclassesandsmalldifferencewithinaclass.Basedonthoseresearchgoalsandrequire-ments,thefollowingcontentwillproposeaninnovativeapproach,whichiscalledSOAW(SmallObjectAsWords),tomodeltheobjectsbydiscussingtheobjectrecognitionmodel.Andfinallythereal-timeobjectrecognitionincomplexenvironmentsisachievedbasedonthestate-of-artmethodsincludingmyinnovativeobjectfeaturedetectionanddescriptionmethodswithintheSOAWmodel.TheexperimentresultsshowthatwithintheSOAWmodel,outtwoexperimentalsystemcanachievethefastlearningandreal-timeobjectrecognitionefficientlyusingtheinnovativeobjectfeatureandcorrespondingfastmachinelearningalgorithms.KEYWORDS:ObjectRecognition,ObjectDetection,ComputerVision,Aug-mentedReality,FeatureDetection,FeatureDescription,Classification-ii-E,|µ¥¢ž8I£O8¹8¹Á‡..................................................................ii=©Á‡.............................................................iii8¹..................................................................vã8¹.............................................................viii1˜ÙXØ.........................................................11.1ïĵ9ÀK¿Â....................................................11.2ƒ'ïÄyGVã.......................................................31.2.18I£O¯K£ã................................................31.2.28I£Oƒ'ïĆ{¤uÐ..................................51.3©z9̇óŠ....................................................61.3.1©z..........................................................61.3.2̇óŠ†SNSü......