slam無(wú)人駕駛傳感器

SimultaneousLocalizationAndSimultaneousLocalizationAnd相機(jī)：?jiǎn)文俊㈦p目激光:超聲波:相機(jī)：?jiǎn)文?、雙目激光:超聲波:[1]Chong,T.J.,Tang,X.J.,Leng,C.H.,Yogeswaran,M.,Ng,O.E.,&Chong,Y.Z.(2015).Sensortechnologiesandsimultaneouslocalizationandmapping(SLAM).ProcediaComputerScience,76,MonoInfraredMonoInfrared（通常方便人或者設(shè)備讀取或者處理（單一通道），（通常方便人或者設(shè)備讀取或者處理（單一通道），CCDArLight 11 Cones:sensitive Cones:sensitivetored,green,andbluelightare63,31,and6percent,respectivelyRods:areexquisitelysensitiveandonlyinlow-lightCIE1931colorImageOpticalPinhole相機(jī)成像（小孔成像模型ImageinvertedSizereducedImageisdimNodirectdepthImageOpticalPinhole相機(jī)成像（小孔成像模型ImageinvertedSizereducedImageisdimNodirectdepthf這就是：perspective假設(shè)給定了世界坐標(biāo)系下的點(diǎn)P,影p(x,Z=Z假設(shè)給定了世界坐標(biāo)系下的點(diǎn)P,影p(x,Z=Z軸平行x//XxfyfPose//Pose//Framecoordinates(xim,yim)pixelsImagecoordinates(x,y)inmmCameracoordinates(X,Y,Z)Worldcoordinates內(nèi)參IntrinsicParameters(ofthecameraandtheframegrabber):linktheframecoordinatesofanimagepointwithitscorrespondingcamera外參Extrinsicparameters:definethelocationandorientationofcameracoordinatesystemwithrespecttotheworldcoordinatepImagecenterDirectionsofaxesPixelsize從3D到PerspectivepImagecenterDirectionsofaxesPixelsize從3D到Perspective(ox,oy):imagecenter(in(sx,sy):effectivesizeofthepixel(inmm)f:focallengthxfyfx(ximox)sxy(yimoy)syPRPwPRPw旋轉(zhuǎn)PRPwPRPwr rYr RT T x xP Tr rYr T RT yr rYr T RT T z z RT 1R RT 3世界坐標(biāo)系到相機(jī)坐標(biāo)系（要在自己的坐標(biāo)系下看RT rYr xP世界坐標(biāo)系到相機(jī)坐標(biāo)系（要在自己的坐標(biāo)系下看RT rYr xPT rYr RTy rYr R z ximoxfxr11Xwr12Yw13Zwr31Xwr32Ywr33Zw o f r21Xwr22Ywr23ZwTy yr31Xwr32Ywr33Zw(x,y)(fX fY x(ximox)sxy(yimoy)sy3D->2D 3D->2D TxMext Ty x1 Xw x2MintMextYwx3 Zw 1 oxMint f oy 1ximoxfxr11Xwr12Yw13Zwr31Xwr32Ywr33Zw o f r21Xwr22Ywr23ZwTy yr31Xwr32Ywr33ZwXwx'Xwx'r32Ywr33Zwr22Ywr23ZwXwXwy' r 32 =xiXir21xiYir22xiZir23xiTyyiXi(r11)yiYi(r12)yiZi(r13)yi(Tx)xiXiv1xiYiv2xiZiv3xiv4yiXiv5yiYiv6yiZiv7yiv8x'(r21Xwr22Ywr23ZwTy)y'(r11Xwr12Ywr13ZwTxx'fy(r21Xwr22Ywr23ZwTy)y'fx(r11Xwr12Ywr13ZwTx幀平面（圖像平面幀平面（圖像平面d=xr-ffd=xr-ffp(x,y OpticalCenterBOpticalCenterZDf2D2D3D表面法向量（SurfaceNormal表面法向量（SurfaceNormal點(diǎn)特征直方圖（PointFeature旋轉(zhuǎn)圖片特征（Spin-迭代最近鄰點(diǎn)（IterativeClosest p+ p+ C1PPPP Cvjjvj,j{0,1,2}0012xp1 12u12unsptpsvu pt wudd pt 2vntuptpsarctanwnt,unt Np Np ER,t xiRpiNp SVD求解矩陣RUVtSVD求解矩陣RUVtER,tx2y22 Npi 0=U 0VT U,VR3x3 Nx Np xi,p Np ER,t xiRpiNp Rusu,RaduBogdan."Semantic3dobjectmapsforeverydaymanipulationinhumanlivingenvironments."KI-KnstlRusu,RaduBogdan."Semantic3dobjectmapsforeverydaymanipulationinhumanlivingenvironments."KI-KnstlicheIntelligenz24.4(2010):345-Johnson,AndrewE.,andMartialHebert."Usingspinimagesforefficientobjectrecognitionincluttered3Dscenes."IEEETransactionsonpatternanalysisandmachineintelligence21.5(1999):433-449.Besl,PaulJ.,andNeilD.McKay."Amethodforregistrationof3-shapes."IEEETransactionsonpatternanalysisandmachineintelligence14.2(1992):239-256.Rusu,RaduBogdan,NicoBlodow,andMichaelBeetz.