基于TOF计算成像的三维信息获取与误差分析

Oct.2018基于TOF计算成像的三维信息获取与误差分析卢纯青，宋玉志，武延鹏，杨孟飞(中国空间技术研究院，北京100081)摘要院TOF三维成像技术具有帧内并行主动探测、获取景物信息的实时性好、环境光影响小、测量精度高、抗运动干扰性强、平均功耗低等优势，在三维智能感知、工业检测、SLAM等领域有着广泛的应用前景，特别是在自主导航,驾驶控制与智能系统中作为获取实时三维成像信息的传感器得到了迅速发展。研究了两类TOF成像的原理及特点，论述了其系统组成，并将TOF成像原理与其他主流三维成像技术进行了对比，对其主要误差来源和类型进行了归类和分析，研究了测量误差模型。TOF成像作为新一代三维成像技术仍处于发展阶段，可以有效提升智能系统的成像感知能力和测量水平，推动了相关领域的技术进步。关键词院TOF；三维场景；幅度调制；相位检测；误差类型中图分类号院O436文献标志码院ADOI院10.3788/IRLA201847.10410043DinformationacquisitionanderroranalysisbasedonTOF

computationalimaging

LuChunqing,SongYuzhi,WuYanpeng,YangMengfei(ChinaAcademyofSpaceTechnology,Beijing100081,China)

Abstract:Timeofflight(TOF)three鄄dimensionalimagingtechnologyhastheadvantagesofactiveparalleldetectionintheframe,goodreal鄄timeacquisitionofsceneinformation,smallinfluenceofambientlight,highaccuracyofdetectiondata,stronganti鄄sportinterferenceandlowaveragepowerconsumption.Inthree鄄dimensionalintelligentsensing,industrialinspection,SLAMandotherfieldshaveawiderangeofapplications,especiallyinautonomousnavigation,drivingcontrolandintelligentsystemsasareal鄄timethree鄄dimensionalimaginginformationofthesensorhasbeenrapidlydeveloped.TheprincipleandcharacteristicsoftwokindsofTOFimagingwerestudied.TwotypesofTOFimagingprincipleandcharacteristics,systemcompositionwerediscussed.TOFimagingsystemwascomparedwithothermainstreamthree鄄dimensionalimagingtechnology.Itsmainsourceoferrorandtypewasclassifiedandanalyzed,itserrormodelwasstudied.Asanewgenerationofthree鄄dimensionalimagingtechnology,TOFisstillinthedevelopmentstage,whichcaneffectivelyenhancetheimagingperceptionandmeasurementleveloftheintelligentsystemandpromotethetechnologicalprogressinrelatedfields.Keywords:timeofflight;3Dscene;amplitudemodulation;phasedetection;errortype收稿日期院2018-05-10曰作者简介院修订日期院2018-06-18袁袁袁尧遥Email:cust0702@sina.com

(1988-)袁1041004-1

红外与激光工程

第10期www.irla.cn第47卷0引言尧尧[2-3]

尧尧遥尧(Time鄄of鄄Flight袁TOF)袁[1]

尧尧遥袁曰袁Q0袁遥Q1遥袁袁院ctp遥袁2袁袁袁遥袁袁袁院Q0Q1tp遥袁850nmLED袁尧袁Z1TOFFig.1TOFimagingsystemdiagram

袁90袁(/遥TOFCMOS袁遥TOF袁尧201.1直接测量法院Z=c窑tTOF

2袁曰tTOF袁(1)袁遥袁(ImpulseTime鄄of鄄FlightImaging)袁CMOS尧TOF)袁遥TOF尧LED(SLAM)尧1TOF成像原理TOF袁TOF袁1袁遥遥尧袁尧遥[4]

LED遥TOFCMOS遥LEDLED850依30nmLEDLED袁TOFLEDLEDQ1袁遥袁Q0遥LEDQ0袁TOFTOFTOF袁1041004-2

红外与激光工程

第10期www.irla.cn第47卷td院td=tp窑Q2Q1+Q2

(2)C2(3)遥袁袁院遥(5)s(t)=A窑cos(t)院曰遥袁袁袁尧曰院TOFTOF院Q1Q2Z院遥C1r(t)=B+kA窑cos(t+院A曰B袁曰k)遥(6)袁Z=c窑td

2袁TOFc()=limT寅肄1T2s(t)乙T2-T2s(t)r(t+)dtTOF(7)院(8)袁Fig.2Directmeasurementlightpulsetiming

r(t)袁1.2间接测量法遥TOF遥袁袁袁(遥院袁TOF)袁袁尧袁(4)袁c()=kAcos(t+2袁4遥袁袁i)+B院TOF(Correlation鄄BasedTime鄄of鄄FlightImaging)袁袁遥袁=t(0=0毅尧1=90毅尧2=180毅尧3=270毅)袁4C(i)(i=1袁2袁3袁4)C(0)=c(0)+B=kAcos(2C(1)=c(1)+B=kAsin(2C(2)=c(2)+B=-kAcos(2C(3)=c(3)+B=-kAsin(2袁袁袁3袁袁袁B袁Popt)+B)+B)+B)+B遥(9)(10)(11)(12)袁ZTOF窑TOFZ=c4仔遥袁袁曰TOF袁院c遥遥TOF4ADCC(i)袁kA1041004-3

红外与激光工程

第10期www.irla.cn第47卷遥院(18)窑TOFZ=c4仔m驻Z=3Fig.3Schematicdiagramofindirectmeasurementmethod

c4仔曰cm驻袁院TOFTOFm驻c(19)m+4仔袁(13)(14)(15)院c袁遥曰驻TOF驻t袁B遥袁TOF遥袁袁TOFTOF2.2景物视场与分辨率TOF遥TOFCCD袁院(20)曰d袁院(21)遥袁kA尧院22kA=姨(C3-C1)-(C2-C0)21024伊1024曰B=1移i=1Ci

44

TOF480伊360[5-6]遥袁=arctanC3-C1C2-C0

2048伊2048遥TOF2TOF成像的主要特点袁袁TOF袁袁3遥遥院(16)(17)袁袁袁Z袁c遥遥SAPD遥D=LdfS袁院f遥曰LS2.1景物深度与分辨率TOFdS=tan+f320伊240袁袁Z=c窑tTOF2院驻Z=tTOF窑c驻tTOF

22尧院驻c遥尧遥尧m蓸蔀L4TOF袁15滋m袁6mm袁70毅伊60毅遥袁曰驻tTOF袁驻tTOF4TOFTOFFig.4CurveofTOFcamerapointclouddistanceandmeasurement

distance

1041004-4

红外与激光工程

第10期www.irla.cn第47卷5袁袁遥院遥袁遥遥袁袁遥袁遥袁袁袁袁袁1袁遥70毅伊60毅10m袁25cm袁TOFTOF遥遥2.3三类成像测量的比较院尧TOF[1,7]

遥63遥遥院曰(1)袁(2)3袁袁袁袁5Fig.5Stereovisionmeasurementschematic

遥表13种主流三维成像技术性能对比Tab.1Comparisonofthreemainstream3DimagingtechniquesCharactersAlgorithmcomplexityHardwarecost

6SizeMeasurement

rate

PassivestereoscopicapproachesComplexLowMiddleMiddleDependonbaseline

BadGood(Noshadow)Low

StructuredlightComplexMiddleMiddleLowHighGoodMiddleMiddle

Timeofflight

SimpleMiddleSmallFastHighGoodMiddleLow

Fig.6Structuredlightmeasurementschematic

尧袁袁袁袁)袁遥遥院(PrecisionAdaptabilityinlowlightAdaptabilityinbrightlightPowerconsumption

(1)尧3TOF成像测量误差源分析TOF(2)1041004-5

红外与激光工程

第10期www.irla.cn第47卷[8]

遥尧尧袁袁袁遥袁遥遥(22)袁遥遥遥袁袁遥袁8TOF9袁3.1成像系统误差TOF院遥TOFLEDTOF[10]

TOF袁7遥TOF袁袁遥袁8袁遥袁LED遥袁袁遥TOFTOF袁袁袁LEDLED艿TOF院c窑A+B4姨2仔fCdA曰B曰Cd遥7院c曰ACd院曰f(23)Fig.7Schematicofmultipatherrorexperiment

Cd=Lmax-LminLmax+Lmin

袁遥遥袁A袁遥TOF遥尧B袁TOF3.2场景误差TOF袁[9]

袁8Fig.8Imagingresultsonarightanglewithdifferentangles

袁遥遥A邑12d袁A遥d院袁袁(24)45毅院90毅袁9袁10毅尧15毅尧30毅遥15毅尧30毅尧45毅TOF1041004-6

红外与激光工程

第10期www.irla.cn第47卷遥袁遥袁袁尧尧遥袁袁袁尧袁尧遥尧3尧遥尧TOF尧遥遥袁遥袁袁TOF参考文献院[1][2]

ZanuttighP,MarinG,DalMuttoC,etal.Time鄄of鄄flightandStructuredLightDepthCameras[M].Heidelberg:Springer,2016.Wasenm俟llerO,AnsariMD,StrickerD.DNA-SLAM:densenoiseawareSLAMforTOFRGB-Dcameras[C]//AsianConferenceonComputerVision,2016:613-629.[3]

HewittRA,MarshallJA.Towardsintensity鄄augmentedSLAMwithLiDARandTOFsensors[C]//InternationalConferenceonIntelligentRobotsandSystems,2015:1956-1961.[4]

GuptaM,AgrawalA,VeeraraghavanA,etal.Apracticalapproachto3Dscanninginthepresenceofinterreflections,subsurfacescatteringanddefocus[J].InternationalJournalofComputerVision,2013,102(1-3):33-55.[5]

KimW,YibingW,OvsiannikovI,etal.A1.5MpixelRGBZCMOSimagesensorforsimultaneouscolorandrangeimagecapture[C]//Solid鄄StateCircuitsConferenceDigestofTechnicalPapers392-394.[6]

(ISSCC),2012IEEEInternational,2012:

9PayneA,DanielA,MehtaA,etal.7.6a512伊424CMOS3Dtime鄄of鄄flightimagesensorwithmulti鄄frequencyphoto鄄demodulationupto130MHzand2GS/sADC[C]//Solid鄄State

Circuits

Conference

Digest

of

Technical

Papers

(ISSCC),2014IEEEInternational,2014:134-135.

Fig.9Angleformedbythetwoplanesimagingresults

4结论与展望袁袁尧袁尧,袁,,袁袁袁1041004-7

袁[7]TIToFsensor.Time-of-flight(TOF)sensors[EB/OL].

[2016-04-15].https://www.ti.com/sensing-products/optical-sensors/3d-time-of-flight/overview.html.[8]

PayneAD,DorringtonAA,CreeMJ.Illuminationwaveformoptimizationfortime鄄of鄄flightrangeimagingcameras[C]//Videometrics,RangeImaging,andApplications

[9]

XI.InternationalSocietyforOpticsandPhotonics,2011,8085:80850D.

StephanM,RahulN,DanielK.Simulationoftime鄄of鄄flightsensorsusingglobalillumination[C]//Vision,Modeling&Visualization,2013:33-40.

[10]MutnyM,NairR,GottfriedJM.Learningthecorrectionfor

multi鄄pathdeviationsintime鄄of鄄flightcameras[J].ComputerScience,2015,1512:04077.

TOF