欢迎来到天天文库
浏览记录
ID:36641744
大小:987.79 KB
页数:77页
时间:2019-05-13
《视频配准技术的研究》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、上海交通大学硕士学位论文视频配准技术的研究姓名:王力申请学位级别:硕士专业:模式识别与智能系统指导教师:周越20081201上海交通大学硕士学位论文AbstractThispaperstudiestheproblemofmatchingtwounsynchronizedvideosequencesofthesamedynamicscene,recordedbydifferentstationaryuncalibratedvideocameras.Thematchingisdonebothintimeandinspace,wherethespatialmatchingcanbemodel
2、edbyafundamentalmatrix.Ourapproachisbasedonmatchingspace-timetrajectoriesofmovingobjects,incontrasttomatchinginterestpoints,asdoneintraditionalfeature-basedstaticimagematchingtechniques.Thesequencesarematchedinspaceandtimebymatchingallpointsalongcorrespondingspace-timetrajectories.Byexploitingth
3、edynamicpropertiesofthesespace-timetrajectories,weobtaintemporalcorrespondencebetweenthetwovideosequences.Furthermore,usingtrajectoriesratherfeature-pointssignificantlyreducesthecomplexityofthespatialpoint-matchingproblemwhenthesearchspaceislarge.Thisbenefitisusedforverywidebase-linematchingwhic
4、hisextremelydifficultwhenonlytraditionalstaticimagematchingisused.Inthispaper,thevideoalignmenttechniqueisinvestigateddeeplyonthebaseofthesummaryofpreviousresearchworksinthisarea.Proposedframeworkinthevideoalignmentincludesfeature-pointsextraction,trackingfeature-points,epipolargeometry,temporal
5、-spatialalignmentbetweentwovideosequences.Meanwhile,weproposeanavailablemethodtoissuetheproblemofverywidebase-linematching.Intheexperiment,themethodproposedinthispaperperformswellincontrasttotheSIFTalgorithm.Thispapersetsthevideoalignmenttechniqueasthemainresearch.Ourworkmainlyincludes:1.Thispap
6、erusesmean-shiftalgorithmtotrackfeature-points.Asaresult,weobtaintrajectoriesoffeature-points.Weusethedynamicinformationintrajectoriestomatchtwounsynchronizedvideosequences.2.Thispaperutilizesepipolargeometrytosolvetheproblemofwidebase-linematchingbetweentwovideosequences.IV上海交通大学硕士学位论文3.Thispap
7、ercombinesthedynamictimewarpingalgorithmandthefundamentalmatrixtoexploretemporalinformationinvideosequences.4.Giventhetemporalcorrespondencebetweentwovideosequences,thispaperfindthespatialcorrespondencebypointsmatchinginallt
此文档下载收益归作者所有