The Eight-Point Algorithm

The Eight-Point Algorithm

ID:40385598

大小:2.34 MB

页数:7页

时间:2019-08-01

The Eight-Point Algorithm _第1页
The Eight-Point Algorithm _第2页
The Eight-Point Algorithm _第3页
The Eight-Point Algorithm _第4页
The Eight-Point Algorithm _第5页
资源描述:

《The Eight-Point Algorithm 》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、RobertCollinsRobertCollinsCSE486,PennStateCSE486,PennStateReminder:-0.00310695-0.00256462.96584F=-0.028094-0.0077162156.381313.1905-29.2007-9999.79Lecture20:TheEight-PointAlgorithmReadingsT&V7.3and7.4RobertCollinsRobertCollinsCSE486,PennStateEssential/Fundamen

2、talMatrixCSE486,PennStateE/FMatrixSummaryTheessentialandfundamentalmatricesare3x3matricesLonguet-Higginsequationthat“encode”theepipolargeometryoftwoviews.Motivation:Givenapointinoneimage,multiplyingEpipolarlines:bytheessential/fundamentalmatrixwilltelluswhiche

3、pipolarlinetosearchalonginthesecondview.Epipoles:EvsF:Eworksinfilmcoords(calibratedcameras)Fworksinpixelcoords(uncalibratedcameras)RobertCollinsRobertCollinsCSE486,PennStateComputingFfromPointMatchesCSE486,PennStateComputingF•Assumethatyouhavemcorrespondences•

4、Eachcorrespondencesatisfies:•Fisa3x3matrix(9entries)•SetupaHOMOGENEOUSlinearsystemwith9unknowns1RobertCollinsRobertCollinsCSE486,PennStateComputingFCSE486,PennStateComputingFGivenmpointcorrespondences…Think:howmanypointsdoweneed?RobertCollinsRobertCollinsCSE48

5、6,PennStateHowManyPoints?CSE486,PennStateSolvingHomogeneousSystemsSelf-studyUnlikeahomography,whereeachpointcorrespondenceAssumethatweneedthenontrivialsolutionof:contributestwoconstraints(rowsinthelinearsystemofequations),forestimatingtheessential/fundamentalm

6、atrix,eachpointonlycontributesoneconstraint(row).[becausetheLonguet-Higgins/Epipolarconstraintisascalareqn.]withmequationsandnunknowns,m>=n–1andrank(A)=n-1Thusneedatleast8points.Hence:TheEightPointalgorithm!Sincethenormofxisarbitrary,wewilllookforasolutionwith

7、norm

8、

9、x

10、

11、=1RobertCollinsRobertCollinsCSE486,PennStateLeastSquaresolutionCSE486,PennStateOptimizationwithconstraintsSelf-studySelf-studyWewantAxascloseto0aspossibleand

12、

13、x

14、

15、=1:Definethefollowingcost:ThiscostiscalledtheLAGRANGIANcostandλiscalledtheLAGRANGIANmulti

16、plierTheLagrangianincorporatestheconstraintsintothecostfunctionbyintroducingextravariables.2RobertCollinsRobertCollinsCSE486,PennStateOptimizationwithconstraintsCSE486,PennStateOpt

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。