Digital Three-Dimensional Imaging in the Infrared at the National Research Council of Canada

Texte intégral

(1)

READ THESE TERMS AND CONDITIONS CAREFULLY BEFORE USING THIS WEBSITE.

https://nrc-publications.canada.ca/eng/copyright

Vous avez des questions? Nous pouvons vous aider. Pour communiquer directement avec un auteur, consultez la

première page de la revue dans laquelle son article a été publié afin de trouver ses coordonnées. Si vous n’arrivez

pas à les repérer, communiquez avec nous à [email protected]

Questions? Contact the NRC Publications Archive team at

[email protected] If you wish to email the authors directly, please see the

first page of the publication for their contact information.

NRC Publications Archive

Archives des publications du CNRC

This publication could be one of several versions: author’s original, accepted manuscript or the publisher’s version. /

La version de cette publication peut être l’une des suivantes : la version prépublication de l’auteur, la version

acceptée du manuscrit ou la version de l’éditeur.

Access and use of this website and the material on it are subject to the Terms and Conditions set forth at

Digital Three-Dimensional Imaging in the Infrared at the National

Research Council of Canada

Beraldin, Jean-Angelo; Rioux, Marc; Blais, François; Couvillon, R.A.

https://publications-cnrc.canada.ca/fra/droits

L’accès à ce site Web et l’utilisation de son contenu sont assujettis aux conditions présentées dans le site

LISEZ CES CONDITIONS ATTENTIVEMENT AVANT D’UTILISER CE SITE WEB.

NRC Publications Record / Notice d'Archives des publications de CNRC:

https://nrc-publications.canada.ca/eng/view/object/?id=42fb3983-2eae-4300-870a-dbc64bcc95e3

https://publications-cnrc.canada.ca/fra/voir/objet/?id=42fb3983-2eae-4300-870a-dbc64bcc95e3

(2)

National Research

Council Canada

Institute for

Information Technology

Conseil national

de recherches Canada

Institut de technologie

de l'information

Digital Three-Dimensional Imaging in the

Infrared at the National Research Council

of Canada *

Beraldin, J.-A., Rioux, M., Blais, F., Couvillon, R.A.

July 1994

* published in the SPIE Proceedings, International Symposium on Optics,

Imaging, and Instrumentation: Infrared Technology XX, Session: IR in

Canada part 2. San Diego, California, USA. July 24-29, 1994. Vol. 2269.

pp. 208-225. NRC 37143.

Copyright 1994 by

National Research Council of Canada

Permission is granted to quote short excerpts and to reproduce figures and tables

from this report, provided that the source of such material is fully acknowledged.

(3)

; 1 x,y,z 12 3 1 4 5 Abstract Key words: :  :  :  :  :  :  :  :  :  :  :  

NRC37143{SPIE{Vol.2269 InfraredTechnologyXX(1994),pp.208{225.

Research Council of Canada

1 Introduction

1

J.-A.Beraldin,M.Rioux,F.Blais,andR.A. Couvillon InstituteforInformationTechnology

NationalResearch CouncilofCanada Ottawa,CanadaK1A 0R6

This pap erpresents work p erformedatthe NationalResearch CouncilofCanada ininfrared rangeimaging,i.e., from 15 m to 18 m. This region of the sp ectrum is chosen for eye-safety reasons. Basic concepts explaining the triangulation principleused inNRC prototyp es arepresented . The requirementsfor laser source andoptics are describ ed insomedetail. Lasersp otp ositiondetectionisreviewedinthecontextofinfraredrangeimagingwithactual designexamplesand detailed calculationsof signal-to-noiseratios. Thesecalculationsare usefulinthe earlystage of adesign. Exp erimentalresultsshowrangeimagestakenwiththe rstprototyp ebuiltattheInstituteforInformation Technology. Adiscussiononcurrentdevelopmentsconcentrates onanotherprototyp erangecameraintendedforspace applications. Thecurrentversionofthatprototyp eop erates at082 mandcanp erformtasksintrackingmo deata refresh rateof130Hz(60targetsp ersecond)orinimagingmo deatadatarateof18000registered3-Dandintensity p ointsp ersecond. Theusefulrangeisab out0.5mto10m.Fordistancesgreaterthan10m,atime-of- ightunitalong with apulsed-laser source op eratingat 154 mwillb e included. Thechange ofop erating wavelengthwillprovidea systemthatiseye-safe andincreasethesignalto backgroundlightrejection forspaceapplications.

Rangeimaging,infrared,tracking,time-of- ight,synchronized scanning,p ositiondetectors.

A numb erof techniques areavailablefor theoptical measurementof the shap eof thesurface of anobject. One can distinguishtwomaincategories: feature-based and structured-lightprojectionapproaches. Inthe feature-based approach, 3-D measurements areobtained forsp eci c areas of anobject where rapid changes ofilluminationo ccur. Thismetho dleadstosparsedatarepresentationofasurface. Withastructured-lightapproach,aknownlightpattern is projected onto ascene such that anarti cial change ofilluminationis imp osed. Three-dimensionalmeasurements forevery pixelintheimageareobtainedbyscanningthelightpatternoverthedesired eldofview. Theresult isan elevationmap(range map)ofthe scene. This approach isfavoredfortheinsp ectionof smo othsurfacesfor which no naturalfeatures areavailableforamatchingalgorithm.

Among the many structured-light (non-contact) techniques available to extract 3-D information from a scene, activetriangulationisusedinapplicationsasdiverseasmeasurementandrepro ductionofobjects,insp ectionofprinted circuit b oards,andautomaticrob otwelding . An innovativeapproach,basedontriangulationusingasynchronized scanningscheme,was intro duced byRioux to allowverylarge eldsofviewwithsmalltriangulationangles,without compromisingonprecision. A3-Dsurfacemapiscapturedbyscanningalaserb eamontoascenealongtwoorthogonal directions andcollectingthere ecte d laser lightfromthescene ontoap ositiondetector. Geometriccorrection ofthe rawdata (lasersp otp osition andangular p ositionsof thescanningmirrors)givestwoimagesinp erfect registration: one with co ordinates and a second with intensity data representing the collected laser p ower from the scene. Thelaserb eamcanb e ofasinglewavelength(visibleorinfrared)or comp osedofmultiplevisiblewavelengthsforthe purp oseofmeasuringthecolormapofascene. Mostoftheprototyp esthathaveb eenbuiltatourInstituteusesingle wavelength lasers , i.e.,spanning the sp ectrum from0633 mto 154 m. One of them can b e mo di edto op erate withawhitelaser. Thisparticularsystemwasdemonstratedformuseumapplications .

This pap erfo cuses on thedesignasp ects ofarangecamera basedon triangulationthatuses asingle wavelength lasersource intheshort wavelengthinfrared(SWIR),i.e.,15 mto18 m,regionofthesp ectrum. ThetermSWIR willb eusedthroughoutthispap ertosp ecifythisregionofthesp ectrum. Someauthorsdistinguishthenear-IRasb eing theregioncovered by07 mto 11 mand theshort-wave infrared(SWIR) from11 mto 3 m. This lab oratory

(4)

 z z p p 6 0 0 0 0 7 2 0 0 x x 0

2 Active Triangulation Principles

2.1 Conventional Active Triangulation

 :  :  :  :  x;z z df p f  x z  p  d f f z  p  z f d   p z f d d P f P

inthe AmericanNationalStandard fortheSafe useof Lasers and dep endingonop erating conditions,systems that uselasersthat emitinthatregionofthesp ectrumaremoreeye safethanthosethatemitintherestofthesp ectrum, i.e.,from0.18 mto1mm. Oneshouldnotethatb etween 15 mand18 mopticalmaterialsusedinthevisibleare still adequate and that delivery ofthe laser b eambyalowloss silicaoptical bre isp ossible. Furthermore, another advantage,particularly forspace applications,is that thesolar sp ectral irradianceis lower intheSWIR than inthe visible.

The opticalarrangementof anauto-synchronized rangecamerais presented inSection 2. The advantagesof the auto-synchronizedscanningschemeoverconventionaltriangulationarehighlighted. Thelasersp otp ositiondetection pro cess isreviewed indetailinSection 3fortwotyp esofp osition detectors that cansatisfydi erentsystem require-ments. Section4addressesdesignasp ects coveringthelasersourceandoptics. Exp erimentalresults obtainedfromthe rstNRCrangecameraop eratingat15 marepresentedinSection 5. AdiscussionfollowsinSection6onacurrent developmentofarange cameraintendedfor space applicationsthat willcombinea time-of- ightunitfor long-range measurementsandapulsedlasersource emittingat154 m. Conclusionsapp ear inSection7.

To help explain the optical arrangement used in the prototyp es describ ed later in the pap er, conventional active triangulationisreviewedandcomparedwithsynchronizedscanning. Theimplementationofsynchronized scanningas anauto-synchronized cameraisshowntohavemanyadvantagesoverconventionaltriangulation.Alsocoveredarethe Scheimp ugcondition,di ractione ects, andrangenoise dueto lasersp eckle.

Thebasic geometricalprincipleof opticaltriangulationisshown inFig.1a. Thelightb eamgeneratedby thelaseris de ectedbyamirrorandscannedontheobject. Acamera,comp osedofalensandap ositionsensitive photo detector, measuresthelo cationoftheimageoftheilluminatedp ointontheobject. Bysimpletrigonometry,the co ordinates oftheilluminatedp ointontheobject arecalculated. FromFig.1a,

=

+ tan( )

(1)

and

= tan( ) (2)

where isthep ositionoftheimagedsp otonthep ositiondetector, isthede ectionangleofthelaserb eam, isthe separationb etweenthelensandthelasersource,and isthee ectivedistanceb etweenthep ositiondetectorandthe lens. isrelatedtothefo callengthofthelens.

Toshowsome ofthe limitationsof this triangulationmetho d, let us approximatethe standard deviation of the error in , ,asafunctionof only . Thelawof propagationof errorsgives

= (3)

where is thestandard deviation of theerror in themeasurementof . Theerror inthe estimate of is therefore inversely prop ortional to b oth theseparation b etween the laser and the p osition detector and the e ective p osition of thelens,but directly prop ortionalto thesquareofthe distance. Unfortunately, and cannot b emade aslarge as desired. islimitedmainlybythemechanicalstructure oftheoptical setupand byshadow e ects. Furthermore, inthe conventionaltriangulation geometry,the eldof view8 of thesensor, assuming thelaserb eamcan scan the whole eldofview,isgivenapproximatelyby

8 2arctan ( 2

) (4)

where isthelengthof thep ositiondetector.

Therefore, inthe conventionalsetup, acompromiseamong eld of view, precision of the3-D measurement, and shadowe ects mustb e considered. Asynchronized geometryprovidesawaytoalleviatethesetradeo s.

(5)

Laser

Object

Deflector

Position

Detector

P

p

X

Z

Lens

d

Mirror

(X,Z)

θ

Φ

x

f

o

θ

θ

Object

Z

Laser

Position

Detector

Lens

(X,Z)

p

X

f

o

CCD

Lens

Fixed

Mirror

Scanning

Mirror

Source

Fixed

Mirror

Projection

Axis

Collection

Axis

Detection

Axis

β

f

f

f

o

γ

CCD

d

3 8 9 2.2 Auto-Synchronized Scanning

Basicprinciple ofactivetriangulation: (a)conventional triangulation,(b) synchronized scanner approach.

Auto-synchronized scanner approach: (a)withdouble-coated mirror,(b)Scheimp ug geometry.

(a) (b)

Figure1:

(a) (b)

Figure2:

Rioux intro ducedasynchronizedscanningscheme,withwhichlarge eldsofviewwithsmalltriangulationanglescan b eobtainedwithoutsacri cingprecision. Withsmallertriangulationangles,areductionofshadowe ectsisinherently achieved. The intent is to synchronize the projection of the laser sp ot with its detection. As depicted in Fig. 1b, theinstantaneous eld ofviewofthep ositiondetector followsthesp otas itscans thescene. Thefo callengthofthe lensistherefore relatedonlytothedesireddepth of eld ormeasurementrange. Implementationofthistriangulation techniquebyanauto-synchronizedscannerapproachallowsaconsiderablereductionintheopticalheadsizecompared to conventionaltriangulationmetho ds(Fig.2a).

Figure 3 depicts schematically the basic comp onents of two typ es of dual-axis auto-synchronized cameras. In Fig.3a, 3-Dsurface mapsareobtainedbyusing twooscillatingmirrorsto scanalaser b eamontoascene, collecting the lightthat is scattered by the scene insynchronism with the projection mirrors, and nally, fo cusing this light ontoalinearp ositiondetector, e.g.,photo dio dearray,charge-coupled device,or laterale ectphoto dio de. Theimage acquisitionpro cess yieldsthreequantitiesp ersamplinginterval: twoarefortheangularp ositionofthemirrorsandone forthep ositionofthelasersp otonthep ositiondetector . ForthearrangementdepictedinFig.3b,thesynchronization is achieved with two opp osingfacets of ap olygonalmirror. This kindof arrangementis used mainly inhigh-sp eed systems,e.g.,videodatarates(ab out10mega-samplesp ersecond) . Thep ositiondetectorusuallychosenforthistask isalaterale ect photo dio de.

(6)

0.5 m

Position

Detector

Laser

Y-Axis

Scanner

X-Axis

Scanner

Collecting

Lens

y

x

z

Galvanometer

Scanner

Position

Detector

Polygon

Scanner

Source

0   ; = = 0 1011 12 13 0 2 0 2 1 2 0 2 2 0 0 1 2 f f fl = l f ! z ! z ! ! z z  ! =e a a !  ! a  2.3 Scheimp ug geometry

2.4 Di raction Limit and Speckle

Dual-axissynchronizedscanner: (a)dualgalvanometersarrangement,(b)p olygonandgalvanometerarrangement.

(a) (b)

Figure3:

AsshownonFigs.2band3a,thep ositiondetectoristiltedcomparedtothecollectinglens. Thisopticalarrangement isknownastheScheimp uggeometry. Figure2bshowsschematicallythegeometrywheretheprojectionandcollection axeshaveb eenunfolded. Theparametersonthe gurearethetriangulationangle ,thedetectortiltangle ,thelens fo callength ,and = ( ). Triangulation-basedrange nderscantakeadvantageoftheScheimp ugoptical arrangement . It ischaracterized by the fact that the detector axis, thelens principalplane, and the projection axisallintersect atacommonp oint. One oftheimplicationsofthisoptical arrangementisthat anyp ointalongthe projectionaxis isin fo cus on thep osition detector. This prop erty providesa considerableimprovementindepth of view. Furthermore, with the Scheimp ug arrangement, the lens ap erture can b e kept at maximum, thus reducing sp ecklenoise (nextsection) andimprovingsignalstrength .

For anylaser rangeimagingdevice the3-Dsamplingprop ertiescan b e estimated bytheaccessible numb erof voxels (volumeelements)withinagivencubic volume . Using theGaussianb eampropagationformula,one nds

( ) = 1 +( ) (5)

where ( ) is theradius of thelaser b eamat adistance fromtheb eamwaist lo cationand is the wavelength of thelasersource. Theb eamwaist hasaradiusof andisde ned astheradiusofthe1 irradiancecontour atthe planewherethelaserwavefrontis at. Figure4ashowsthepropagationofalaserb eamwithinthevolumeofinterest. Then,ifthedepthof eld is de nedfollowingtheRayleighcriterion,

= 2 (6)

Suchaconstraintimp osedbythedi ractionlimitleadsto thefollowingrelationship.

= ( 2

(7)

a

a

a

2

ω

0

Resolvable Volume Elements

Depth of Field (m)

0

1

2

3

4

5

6

1000

1500

2000

2500

3000

3500

4000

4500

5000

=   etal. 3.1.1 Op eration  :  V V !  a   :     F F 0 1 2 3 z p 9 p 14 15 7 15 16

3 Laser Spot Position Detection

3.1 Discrete ResponseDetector: Linear Array of Photodiodes

Gaussianb eam: (a)propagationin space,(b)resolvable volumeelementsfor m.

(a) (b)

Figure4: = 154

Themaximumnumb erofresolvablevolumeelements( )alongeach axiswithinthevolumeofinterestis

= 2

= 2

(8)

Figure4bshowsagraphicalrepresentation ofthis functionfor =154 m. Asanexample,withinacub eof3mon aside,onehasaccessto3500 voxels.

Lasersp eckleisanimp ortantissuewithrangecamerasb ecauseinmostcasesitwillimp osealowerlimitontheir range measurement precision . The range precision in a synchronized scanner is in practice limited by the laser sp eckle impinging on the p osition detector. Sp eckle manifests itself on the measurementof the laser sp ot p osition on the detector. Thevalueof willdep end on the typ e of detector that one cho oses fora camera. With current technology,ahigh-sp eedrange nderwilluseacontinuousresp onsedetectorlikealaterale ect photo dio de. Beraldin describ easystemthatusessuch ap ositiondetectortogenerate registeredrangeandintensitydataat10-Mega samples p er second. Thevalueof can,for this kind of system, b e approximated withthe noise mo del presented in forlowcollected p ower. For highcollected p ower,the contributionof lasersp eckle forthis detectoris describ ed in . Usually,range ndersbased onadiscrete-typ e detectorlikeself-scanned photo dio des arraysandCCDs provide data withhigher levelsof accuracythan thosebased on acontinuousresp onse detector but at areduced data rate. Forthisp ositiondetector,thelasersp otp ositionmeasurementuncertaintyhasb eenmo deledin whensp ecklenoise dominates. Bothp ositiondetectors arecoveredinmoredetailinthefollowingsection.

This section examines in more detail the two most common p osition detectors used to extract the lo cation of the lasersp otinanactiverangecamera,i.e.,discrete andcontinuousresp onsedetectors. Theformerisimplementedwith a lineararray of photo dio des and the latter with alateral e ect photo dio de. The op eration, the sp eckle noise, the bandwidthrequirement,thedynamicrange(noiselevel)andadesignexamplearesuppliedforeachdetector.

Recent advances inInGaAs materialshave made p ossible the fabrication of commercial SWIR detector arrays that op erateatro omtemp erature . Commercialdetectorarraysarecomp osedofanarrayofphoto dio desmadeofInGaAs

(8)

Phase 1

Phase 2

Start

Shift Register

End of Scan

+

Switching

Transistors

Bonding

Pads

InGaAs

Photodiodes

Substrate

R f

C f

C v

C

D

Laser

Peak Position

Image of the Sun

Laser Pulse

17

18

   

Equivalentcircuitof detectorarray.

Lasersp otdetected even withpartlysaturateddetector. Figure5:

Figure6:

thatarewire-b ondedtoaMOSFET-based multiplexermadeofsilicon. Thewidthoftheactiveareaofonephoto dio de variesfrom25 mtoab out50 mandheightfrom100 mtoab out500 m. Arraylengthsvaryb etween64and1024 photo dio des. These arraysop erate ontheprincipleof chargingthe photo dio desjunctioncapacitance withasuitable bias voltage present on the videoline during the sequential read-out and letting the impinginglight discharge (by photon-generated carriers)thejunctioncapacitance. Thisvideolineconnectsallthephoto dio desthroughaMOSFET (as shown on Fig. 5). Finally, an external pre-ampli er p erforms the charge-to-voltage conversion. Newer arrays combineanampli erwitheach photo dio de . Theiroutputis thenmultiplexed. Themaximumclo cksp eed isab out 5MHz. Theoutputofthedetectorisastreamofpulseswhere theenvelopis afunctionofthelasersp otdistribution. The timeo ccurrence of the p eakonthe videosignalis directly relatedto the geometricalp osition of thelaser sp ot. For example,a 512-photo dio dearray clo cked at 5MHz yields amaximumsp otp osition measurementrateof ab out 9.76kHz,ifitisassumedthatthelaserp ositioncanb eextracted atapixelrateof5MHz. Thecircuitdescrib edb elow wassp eci callydesignedforthat purp ose.

Sub-pixelinterp olationisp erformedbyadigital niteimpulseresp onse(FIR) ltercontainingaderivativestage . Theoutput ofthis lterisused tointerp olatethezero-crossing ofthesignal. A validationstep isaddedto eliminate falsedetections. Maximump eakp ositionextractionrateisachievedbyallowingtheimagedlasersp ottoextendovera numb erofphoto dio desandinterp olation. Theop eratorprovidesgo o dattenuationoflowandhighfrequencynoiseand esp ecially ofthefrequency comp onentsinducedbytheclo cksused toaddress andreadeach photo dio de ofthearray. Immunitytointerferinglightsources (arti cialor natural,e.g.,thesun) or even frommultiple re ection ofthe laser source itselfiscriticalfor achievingarobustp eakp ositiondetector. Themetho dimplementedhere foralineararray of photo dio descan stillop erate even with apartlysaturated sectionon it, as shownon Fig.6. Section 4.3 provides moreinformationontheo ccurrence ofthiscondition.

(9)

p = = = etal. p p  p  0 s 12 p 0 0 0 o p 11 12 p p 0 w s s w 0 s s w 0 2 w Array T e 2 T 1 2 T e 2 T 1 2 T bi r 1 2 j0  p  f D   f D f D  :   :   p    :  !  p  B  f p ! B f p ! =e p  D R Q =q n Q q n Q V V C 3.1.2 Sp eckle 3.1.3 BandwidthRequirement

3.1.4 Dynamic Rangeand NoiseSources

Bandwidthrequirement fora512-elementInGaAsarrayclo cked at5MHzand withapixelwidth of50 m. 100 1.33 150 0.83 200 0.66 250 0.53 300 0.44 Table1:

Barib eau andRioux predicted that sp ecklenoiseforsuch ap ositiondetectorandp eak ndingmetho db ehaveslike aGaussianpro cess andtheestimatedrms uctuationof determinedbysuch noiseisgivenby

= 1 2 cos ( ) 1 (9)

where isthewavelengthofthelasersource, isthee ectivep ositionoftheimaginglens, isthelensdiameter,and isthetiltangleofthep ositiondetector. Letustakeadesignexample: with =110mm, =20mm, =154 m, and =55 ,one gets =589 mor1/8.5ofdetectorpixel(pixelwidth=50 m).

Toaugmenttheprecision on ,one cannotstrictly averageanumb erofsp otp ositionmeasurements,b ecausethe imagedsp eckle patterndo es notchangewhen multiple samplesofthesamesurface patch aremeasured. For agiven surface patch themacroscopicstructure do es notchange. Spatialaveragingistheonlysolutionto reduce . Dremel prop oseintro ductionofspatialaveragingintheobjectplaneorinthepupilplaneoftheimaginglens. The rst metho disinherenttothesynchronizedscanner. Whenaconstantdepthsurfaceisscanned,theimageofthelasersp ot remainsalmoststationary onthep ositiondetector . Because alineararrayof photo dio desresp onds totheintegral of thelightpatternduringameasurementinterval,thenetresult isafour-foldimprovementontheprecision , i.e.,

=147 mor1/34ofapixel. Forother surfaceorientations,thisfactorvariesb etween1and4.

Priortodigitizingthelasersp ot,ananti-aliasing lterwithlinearphaseisnecessary. Thechoiceofthe lterbandwidth isatradeo amongnoisebandwidth,aliasedcomp onents,andaccurate repro ductionoftheenvelop eofthelasersp ot. Thecriterionusedisthat the ltershouldintro ducelessthan1dBattenuationatthecuto frequencyforaminimum sp otsize of =100 m( =50 m). Hence, thebandwidthrequirementofthedetection chain(charge-to-voltage converterand anti-aliasing lter) isgivenapproximatelyby

2 ln2

(10)

where is therequired-3dBsystembandwidth, isthepixelclo ckrate, is thepixel width,and isthesp ot radius at 1 . The envelop e of the laser sp ot is mo deled by a Gaussian distribution. According to exp erimental measurements,thismo delisfoundsucientfordesignpurp oses. Table1showsthebandwidthrequirementfortypical sp otsizeswithacommercial512-elementInGaAsarraywith =50 m.

Thedynamicrangeofanarrayofphoto dio desisde nedas

= ( )

(11)

where isthetotalequivalentchargecollectedonaphoto dio de(saturationcharge), istheelectronic chargeand ( ) istheequivalentinputreferrednoise inelectrons. Thesaturationchargeofapixel,assumingthecapacitance ofthejunctionvarieswiththereverse voltageacrossit,isgivenby

(10)

0 3.1.5 Example        R R 0   2 f n f n a f f a a a a f f f n i i q N f q kTC C e q i P i N P C B C e B i R kTR B e B B B e i e R C B B B R C C C C T k f p N bi r j0 19 19 20 21 22 2 T ph d e s 2 e T 2 2 2 e ph det d det o T 20 s 2 s 2 2 1 2 3 s 3 1 2 1 2 2 2 1 s 1 v D 1 v D 7 19 5 17 junctioncapacitanceat zerobias .

Withlineararraysofphoto dio des,onecandistinguishtwosourcesofinterference: (1)thosethatarestationaryfor a xedtemp erature andclo ckrateand(2) thosethat arerandominnature. Theformer isoftencalled xed-pattern noise andincludesamongotherthings

Darkcurrentvariabilityacross thearray,

Sensitivityvariationb etween photo dio des,

Readingclo ckfeedthrough.

Randomnoise sources arecomp osedof

Photonshot noiseoftheincidentlight,

Shotnoiseofthedarkcurrent,

Resetnoiseofthepixelsinthearray,

Currentandvoltagenoise ofthepre-ampli er.

Twoothersources ofnoise areidenti edintheliterature: theampli erglownoise andp op corn noise . These two sources ofnoisearenotcovered here.

Theequivalentinputreferrednoise (inelectrons)duetotherandomnoise sourcesis

=

( + ) +

1

+ (13)

where = is the incident light(laser and background), is the averaged darkcurrent, is the numb er of pixels (photo dio des), is laser p ower collected on one photo dio de b efore saturationis reached, and is the photo dio de resp onsivityat thewavelengthofinterest. Thedarkcurrent hasastrongdep endence ontemp erature. It is halved for every decrease of 8 C. is thetotal capacitance that is reset to the video-line bias. It includes the photo dio de junction and parasitic capacitance and the video-lineto clo ck-lines capacitance . The integrated noise (over )dueto thechargeampli erofFig.5isfoundbycomputingthevoltagenoise attheoutputoftheampli er andthenwiththegainfactor( )convertingfromvoltagetocharge,i.e.,

= ( + 4 )+ + (

3

) (14)

where and are,resp ectively,theampli ercurrentandvoltagenoisesp ectraldensities, and arethefeedback resistance andcapacitance,resp ectively, and isthefrequencyof thezero intro duced inthenoisegainofthe ampli er, (2 ( + )) . and arethevideolineandphoto dio decapacitances, resp ectively. The temp eratureisgivenby and isBoltzmann'sconstant. Usually,these noisesourcesinclude ickernoise (noisewith 1/ typ edistribution). Here,itisomittedb ecausethederivativetermintheFIR lterofthep eakdetectoreliminates that noise comp onent. Furthermore, xed-pattern noise should b e removedfor the di erent op eratingconditions of thearray,e.g.,temp eratureandintegrationtime,toimproveop erationinlowersignal-to-noiseratios(SNR).Usually, forhighSNR,thee ect oftherandomnoise andthequantizationnoiseofthep eakdetector onthemeasurementof

areswamp edbysp ecklenoise.

Table2presents adesignexampleforanInGaAsphoto dio dearraycomp osedof =512photo dio des,whichisread at 5MHz and isconnected to atypical chargepre-ampli er. Thedynamicrangeis equivalentto 14 or15 bitsof an analog-to-digitalconverter. Converterswith14-bitresolutionarereadilyavailableat5MHz. Acloserlo okatthenoise contribution ofthedi erentsources showsthat 94.5%(ratioofvariances)comes fromthereset noise. Therefore,one cantakeadvantageofatechniquecalledcorrelateddoublesampling(CDS)toreduce substantiallythisnoisesource . Correlated double sampling,however,doubles theampli ernoise p ower comp onent (14). Nevertheless, thedynamic rangewould stillb e ab out2 10 , i.e.,almost18 bits. Correlated doublesamplingisnow integrateddirectly inthe multiplexersusedwithInGaAslineararrays .

(11)

a a pin p p 2 0 2 0 R R

Designexample for512-elementInGaAsarray. s w h 0 r d j f f s 6 0 0 23 2 1 1 2 1 2 ph ph 1 2 det det 14 LEP LEP p j p j f p  p  !  V I C R C i Hz e Hz B ! !  p i i i i ; p ; i i L= L i i i i P P f f  R C R C : 3.2 Continuous Response Detector: Lateral E ect Photodiode

Saturation Charge Saturation Power Noise Charge Noise Power Dynamic Range 3.2.1 Op eration N 512 5MHz Pixelwidth 50 m Pixelheight 100 m Sp ot size( ) 125 m -5V T 300K 8pA 2.5pF 5.6k 2.7pF 0.1 pA/ 5nV/ 1MHz p er pixel 40 10 e for 500nW p erpixel 1600e for 20pW 25000 Table2:

Alaterale ectphoto dio de(LEP)sensitiveto1.54 mhasb eendemonstrated . Figure7illustratesthebasicstructure of a typ eInGaAs/InP single-axisp osition detectoralong withtwotransimp edanceampli ers. Carriers pro duced by lightimpingingon the device areseparated in thedepletion regionand distributed to the twosensing electro des according to Ohm'slaw. Assumingthat alldriftshave b eencanceled inthe electronicsthen thenormalized p osition ofthecentroidoftheincidentlightdistributionisgivenby

= +

[ 1 1] (15)

where and are measuredwithtransimp edanceampli ers. Theactualp ositiononthedetectorisfoundby multi-plyingtheab oveequationby 2where isthedistanceb etween thetwosensing electro des. Thetotalphoto current

is

= + = (16)

where istheincidentlightp owerand istheresp onsivityofthephoto detector,whichisintheorderof0.7A/W.

Thefrequencyresp onseofajunctionphoto dio decanb ea ectedbytwomaintyp esofphenomena: (1)transit-time e ects, which arise fromthe delay b etween the absorption of theincidentlightand the separation of photo-excited electron-hole pairsand (2) theinherent electrical characteristics of thep osition detector, i.e.,theseries resistance of thedio de anditsjunctioncapacitance. Ifthemaximalrequired bandwidthforrange ndingapplications isless than 4MHz,thedistributed resistanceandcapacitance of thelaterale ect photo dio deand thetransimp edanceampli ers dominatethe transientb ehaviorof thep osition detector . One can estimate thebandwidth ofalaterale ect photo dio deas

= 2

1

(17)

where istheinter-electro deresistanceand thejunctioncapacitance. Equation(17)isa rst-orderapproximation and appliesonlywhen theloadimp edancesare equalto zero. An inter-electro de resistance of56kandajunction

(12)

N Layer InP

I Layer InGaAs

P Layer InP

+

+

V

1

R f

C f

I

1

V

2

R f

C f

I

2

L

p

hv

= = p p  0 p

InGaAslaterale ect photo dio de.

p  f D   f D   !  !  F t x x v t  x x v t D R i i i i 23 24 14 12 p 0 0 p 7 0 0 0 0 t 0 s 0 t 0 s Lep sat 2 n 1 2 sat 2 n 1 2 3.2.2 Sp eckle 3.2.3 BandwidthRequirement 3.2.4 Dynamic Range Figure7:

capacitanceof39pFatabiasof-5Vweremeasuredfortheprototyp elaterale ectphoto dio de . Hence,themaximum bandwidth of the present detector is 7.2 MHz, which is quiteadequate for video-rate rangecameras . A complete analysisofthee ectof transimp edanceampli ersonthetransitb ehaviorofthedetectorapp ears in .

Barib eau and Rioux predicted that sp eckle noise for a centroid detector b ehaves like a Gaussian pro cess and the estimatedrms uctuationof determinedbysuchnoiseis givenby

= 1 2 cos ( ) 1 (18)

where isthe wavelengthof thelasersource, is thee ective p ositionof theimaginglens, is thelensdiameter, and isthetiltangleofthep ositiondetector. Accordingtothisequation, foracentroiddetectoris b etterthan onethatusesaFIRderivative lter . But,inpractice,mostsystemsthatop erateathighdatarates ( 100kHz)usea laterale ect photo dio deandtherefore thermaland quantumnoisesources, notsp eckle noise,willlimitthedetection pro cess.

Therequired systembandwidth isdeterminedby thesizeof thelasersp ot andthesp eed atwhich itscans afeature onascene. Itiscustomarytocomputethesystembandwidthorequivalentlytherise timeofthecollectedlightfrom the scene as thelasersp ot movesover atransition inre ectance . Again, one assumesthe lasersp ot distributionon thedetectorto b eGaussian withdimension . Thesp otsize as determinedbydi ractionisrelatedtothestandard deviationofaGaussiandistribution, ,by = 2 . Theresultingnormalizedtimeresp onse waveformisgivenas

() = 1 2 erfc ( ( + ) 2 ) (19)

where isthe p osition of there ectance transition, is theinitialp osition ofthe centroidof the lasersp ot, is laser sp otsp eed onthe scene, and is thetimevariable. Fromthis equation, one can computethe10%to 90%rise timeandhence thesystembandwidthrequirement.

Thedynamicrangeofap ositiondetectorbased onalaterale ectphoto dio deisde ned as

= ( )

(20)

(13)

p p e = = ; "     #    

4.1 Laser Sources for the SWIR

4 Sources of Light and Eye Safety Issues  !  V I C R R i Hz e Hz B i e R R R KTB R R i q B i i e i B R R i R R R :  :   0 r d j p f a a s 6 2 n 2 a f p 2 p 2 s f p 2 a s d ph 2 a 1 2 2 a 1 2 s f p d 925 p p p

Designexample forInGaAs/InPlaterale ect photo dio de. SaturationCurrent Saturation Power NoiseCurrent Noise Power Dynamic Range 3.2.5 Example LEPlength 10mm LEPheight 200 m Sp ot size( ) 50 m -5V T 300K 400nA 39pF 5.6k 220k 0.6 pA/ 4.5 nV/ 200kHz 1mA 1.43mW 1.0nA 1.43nW 10 Table3:

Figure7isusedtocomputethee ectofthedi erentinputnoisesourcesontheoutputsignals. Theresultisgiven b elow = 1 + 1 + 1 + 4 1 + 1 + +2 2 + (21)

where( ) and( ) arethepre-ampli erintegrated(over )inputnoise voltageandcurrentnoise,resp ectively, isthefeedback resistorvalue, isthep ositioningresistor value,and isthedarkcurrent.

Table 3 presents a design example for the InGaAs/InP laterale ect photo dio de built to NRC sp eci cations. The followingdesignparameters were used forthecameradescrib ed inSection 5. Thedynamicrangeisequivalentto 20 bitsofananalog-to-digitalconverter. Converterswith16-bitresolutionareavailableat1MHz. Makinguseofthefull dynamicrangeof anLEP p oses quiteachallenge tosystemdesigners . Acloser lo okat thenoise contribution of thedi erentsourcesshowsthat62%(ratioofvariances)comesfromthethermalnoiseof and28%fromthevoltage noiseoftheampli ers.Therefore,thenoiselevelcouldb ereduced byincreasing andbyselectinganampli erwith a lowervoltagenoise. But,unfortunately, thedynamicrange would notincrease accordinglyb ecause thesaturation currentisalsoafunctionof .

Thissectionsurveysthedi erentlasersourcesfortheSWIR,inparticularfortheregion15 mto18 m. Asummary onmaximump ermissibleexp osuresp ertinenttoeyesafelaserrangecamerasop eratinginthatregionofthesp ectrumis tabulatedandcomparedtoothercommonwavelengths. Theopticaleciencyofarangecameraandthesolarsp ectral irradiancearecovered also.

Twocategoriesofinfraredlasersources thatare quiteadequateforrangecamerasareemerging,i.e.,solid-statelasers andlaserdio des. Numerouspulsedandcontinuouswave(cw)lasersourcesemittingintheSWIR(1.5-3 m)areb eing

(14)

P :Incident

Laser beam

Ac: Collecting

Aperture Surface

Specular

Reflection

Radiant Power

from a Diffuse

or

Lambertian Surface

N

α

: Solid Angle

L

R

;   ' 0 1 2627 28 29 30 det L 31 det L c 2 c 32 o P P P P T   T  R A R A R :  :  D f  z > p z : 4.2 Optical Eciency

4.3 Solar Spectral Irradiance

E ectofsurfacere ectivity onthe received laserp ower. Figure 8:

develop edforapplicationsincludingeye safelaserradarsystemsandmedicalsystems. Solid-statelasershavereceived a lot of attention inrecent years , due in part to a reduction of the cost of high-p ower laser dio des as apump source. Commerciallasers with adequate p ower levelsfor rangecameras have started to app ear onthe market. At theInstituteforInformationTechnology,twotyp esof solidstate lasersweredevelop edforrangingapplications and another foroptical timedomainre ectometers . Laser dio des have also started to breakground inthe high-p ower arena(ab ove300mW) .

The estimationof the required laser p ower is an imp ortant step in the design pro cess. From Fig. 8, the equation relatingthep owerimpingingonthep ositiondetector tothelaserp ower foradi usive surface is

=

cos

(22)

where is thetransmission co ecient accounting for the fo cusing andcollecting optics, is the di use re ectance co ecient,and istheangle b etween thenormalofthesurface b eingmeasuredandthesolid anglesubtended by thecollectingap erture. Thesolidanglecanb e approximatedfor muchlargerthantheap erturediameter by

(23)

where istheareaofthecollectingap ertureand isthedistanceb etween thesurfaceb eingilluminatedbythelaser b eamandthecollectingap erture.

In aspace environment,a range camerawillhave to op erate under ahigh levelof solar irradiance. Therefore, it is imp erative that the camera b e designed or equipp ed with sun blo cking capabilities. A range camera based up on a synchronizedscannertechniquehasaninherentwaytoreducespuriouslight,i.e.,theinstantaneous eldofviewismuch smallerthanthetotal eldofviewofthecamera. Anotherwaytoreducesolarirradianceistouseaninterference lter tuned tothelaserwavelength. Furthermore,ifoneslo oks atthesolarsp ectralirradianceoutside theatmosphere ,a ratioof5.8 isfoundb etween thesolarsp ectralirradianceat0632 mand 154 m. Finally,thelastmetho dofsolar interference rejectionisduetotheScheimp ugcondition(see Section2.3). Figure9showsanexampleofthefo cusing of the sun onap osition detector for =55 , =20mm, and =100 mm. Because of this optical arrangement, the sun,lo catedat in nity, isalwaysfo cusedinthe fo calplane ofthe lens; thereby its imagewillspread outon the p ositiondetector except at theintersectionofthe fo calplaneandthedetector. This last situationcould confusethe p eakdetection pro cess. Foradiscreteresp onse detectorassp eci ed inTable2,thecollected p owerp er nanometreof bandwidth of aninterference lter is25 W (transmission of optical pathis 30%). Hence, at , apixel will collect79.6nW,whichislessthanthesaturationlevelof90nWp erpixel.At corresp ondingto =05,thecollected p oweris25timeslessintense. Itcanb econcludedthat,whenthesunapp earsintheinstantaneous eldofviewofthe

(15)

f

Z

Z = 0.5 m

1 mm

2mm

5 mm

10 mm

p

0 ' ' 2  6 6 5 4 3 2 3 33 0 3 o z L 33    :       !  :  T = :   P 

5 Infrared Range Camera Prototype Imagesizeof thesunona tilteddetector.

4.4 Maximum Permissible Exposure

5.1 Position Detector

Figure9:

camera,thedynamicrangeofthedetectorwillb ereducedforobjectslo catedatlongerdistancesandincreased asthe object getscloser tothecamera.

For applicationsrequiring eye safety, maximump ermissibleexp osure (MPE) isvery imp ortant. It varies with wave-lengthandop eratingconditionsofasystem. TheMPEisde nedastheleveloflaserradiationtowhichap ersonmay b eexp osedwithouthazardouse ectoradversebiologicalchangesintheeyeorskin . Thisstandardalsocontainsmany provisionsforthesafeuseoflasersystemsandforthepreparationoftheenvironmentsurroundinglasersystems. Before attemptinganyexp erimentthatrequireeyesafety,itisrecommendedtoconsulttheANSIstandard andtogetprop er certi cation forcommercialuse ofsuch a lasersystem. For example,fractions of amilliwattinthe visiblesp ectrum aresucienttocauseeye damage,sincethelaserlightenteringthepupilismagni ed,attheretina,throughthelens, byafactorofroughly10 (intra-b eamviewing). Atlongerwavelengths, however,strongabsorptionbywaterpresent intheeye serves toreduceopticalp owerreaching theretina, thusincreasingthedamagethreshold. Waterabsorption isthemechanismbywhich thelaserlightisstopp edat thesurfaceofeithertheskin ortheeye. Fromtheabsorption co ecientofwaterat1.5 m,ane-foldingdistanceof0.6mmisfound,whichmeansthatanydamagethatcouldo ccur willalwaysb esurface e ects andnotdamagetotheretina. FromTable4,at1.54 mtherecommendedsingle-pulse threshold energy foreyedamageisroughly10 timesgreater thanitisatthevisiblewavelength( =0632 m). For longerexp osure, Table5,therecommendedthreshold p owerfactorisdownto 10 . TheANSI standardforeye safety takesintoaccountthatsomelaserlightat 1.54 mistransmittedthoughthecorneaandreachesthelens. Therestis absorb ed bythevitreousb o dyoftheeye.

ConsideringthatthenoiselevelsofInGaAsp ositiondetectors areab outthesameasthatofsilicon-basedp osition detectors and that scattering strengths of rough surfaces are similar at 1.54 m and inthe visible region, there is p otentialforsignalimprovementof 30-40dB(optical)if 1.54 mischosen insteadof thevisible formakingeye-safe range measurements. Noise level of InGaAs p osition detectors was covered in Section 3. Scattering prop erties at 1.54 m forarangeof surfaces includinggo o d di users, e.g.,Al O p owder onglass andon Macor substrates, and, veryadsorptivesurfaces suchas humanskinwerecompared withmeasurementsmadeat0.514 m.

For thisinfrared rangecamera, theaveragesp otsp eed is ab out420m/s. Forasp ot size =250 minthescene, the 10%to 90% rise time isab out 15 s (19). This rise timeis equivalentto abandwidth of 230kHz for alinear phase lterofordersix. Adatarateof600kHzwaschosen forthe12-bitanalogto digitalconverters. Letusassume 50%, =005(skin), 9 10 ,and =45 . Then,forastando distanceof650mmandforanequivalent p owerontheanalog-to-digitalconverterof2 W(for10bitsofprecision ), 15.5mW(22). Thisvalueiswithin reach ofthe lasersource used. It isworthnotingthat rangeprecision forthis cameravariesalmostlinearlywith the collected laserp ower . FromTable3andforacollected p owerof0.2 W,therangeprecisionis ab out7bits.

(16)

0 0 0 0 0 0 0 0 0 0 0 0 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2  2  : : : : :  : : : : :  x y    : : =   2 9 6 9 6 9 6 9 9 6 2 4 4 6 3 4 3 3 4 3 4 4 6 33 6 4 2 4 5.2 The Laser Scanner

5.3 Experimental Results

MPEforOcularExp osure (Intrab eamViewing)toaLaser Beam(single pulsemo de) .

MPEforOcularExp osure (Intrab eamViewing)toaLaser Beam(longp erio d) .

( m) (s) (J/cm ) (mm) 0.632 10 10 050 10 7 0.820 10 10 086 10 7 1.060 10 10 50 10 7 1.540 10 10 10 1 2.000 10 10 01 1 Table4:

Wavelength Exp osure Duration MPE EyeAp erture Diameter

( m) (s) (W/cm ) (mm) 0.632 10 to3 10 1698 10 7 0.820 10 to3 10 0556 10 7 1.060 10 to3 10 16 10 7 1.540 10to3 10 01 3.5 2.000 10to3 10 01 3.5 Table5:

Figure 10ashowsaphotograph ofthe prototyp e that hasb een mo di edtoop erate at 1.5 m. Thescanninghead consists of a six-faceted pyramidalmirror that provides the -axis scanning (see Fig. 3), a at mirror driven by a stepping motorprovidingthe -axisscanning, and projectionand detection units. The followingmo di cationswere alsomade: (1)replacementofthedio delaserbythe brelaseroutputand(2)replacementofthesiliconlaterale ect photo dio debyanInGaAsunit(Section3.2). Nochangeintheopticalelementswasneeded,onlyminoradjustmentsfor thefo cusingofb oththeprojectionanddetectionunitstocomp ensatethee ectoftheindexofrefractionchangewhich o ccursbyshiftingthewavelengthfrom0.820to1.5 m. Thesamplingrateofthep ositionsensorremainedat600kHz usingtheexistinghardware. The1.5 m brelaserdeliversupto80mWthrougha50-mlengthoftelecommunication grade single-mode bre. Of this light, 40 mW is projected into the scene owingto losses in the camera projection optics. Thisamountissucientforthetargetedrangeresolution.

Therecommendedmaximump ermissibleexp osure(MPE) at1.5 mwavelength,consideringthatthelaserb eam sp ends 22 10 simpingingonthepupil oftheeye,is

MPE=10J cm (24)

Toresp ect thisexp osurelimit,theCWscanninglaserp owermustb elessthan35.6Wforarecommendedeyeap erture of0.1 cm. Thisisab out10 larger thanthecorresp ondinglimitcalculatedforanHe-Ne wavelengthof0.633 mand wellb elowthemaximump oweravailablefromcurrent infraredlasersources.

Figure10b-cshowsarecordingonahumansubjectthathasb eentakenwithanincidentp owerlevelof40mW( 900 less thantheMPE).Figure10bisthe1.5 mintensityimageand Fig.10cistheregisteredrangeimage. Agreylevel co ding is used to displayrange information. White is for surface elements close to the camerawhereas black is for elementslo cated atthefurthestextremeofthevolume ofmeasurement. Ontheeye,itisinteresting tonote thatthe waterlayerontheconjunctiva(sclera)isnotthickenough topreventscatteringandgo o dreadingoftheshap eo ccurs up to theedge of thecornea. Image acquisitiontimeisof the order of 1s. The pro les fromeach of the six facets are averagedto decrease thescanningnoise intro duced by thefacet-to-facet angular tolerances. The volumeofview

(17)

2 2 2 2 2 z o o 34 34 6  :  x;y z x y z :    :   :   :   z : z z>   

6 Current Developments: Range Camera for Space Applications Exp erimentalresultsfor m:(a)laserscanner prototyp e,(b)intensityimage,(c)rangeimage.

Figure10: =15

is 250mm 250mm 150mmalongthe , and axes, resp ectively. Thesamplinginthe and axesis ofthe orderof1mmwhereastheresolutionalongthe axisisoftheorderof366 m(1part in4096). Therangeprecision

wasestimated atab out250 monskin.

AphotographofthisrangecameraisshowninFig.11aandaschematicreprese ntationoftheopticalarrangement app earsinFig.3a. Thecameraop eratesineitheravariableresolutionmo deorarastertyp emo deatamaximumdata rate of 18000p ointsp er second. Inthevariableresolution mo de,as illustratedinFig.11b, thelaser scanner tracks targets and geometrical features of objects lo cated withina eld of viewof 30 30 and with arange fromab out 0.5mto100m. Forobjectslo catedatdistancesgreaterthan6m,co op erativetargetsontheirsurfacesarerequiredfor go o dsignal-to-noiseratio. Therangecamerauses twohigh-sp eedgalvanometerstosteerthelaserb eamtoanyspatial lo cationwithinthe eld ofviewofthecamera. Acompactdual-galvanometer controllerwas designed sp eci callyfor thistask. Eachaxiscanb eaddressedbymorethan16bitswithastepresp onseof18 soverwideangleseventhough themirrorsarequitelarge. Thisincrease insp eed andversatilityovertheoriginalmanufacturedcontrollerallowsthe generationofLissajouspatternsfortrackingofobjectsatarefreshrateofupto130Hz(60targetsp ersecond) . The raster mo deis used primarilyfor the measurementof registered range and intensity informationof large stationary objects . Figure12showsanimageofaquarter-scale mo del(lo catedattheNationalResearch Council)ofthecargo bayoftheSpace ShuttleOrbiter. Thescalemo delmeasures4.33mby1.42mby0.6m. Although,itwasdigitizedto aspatialresolutionof2048 4096,the gure displaysare-interp olatedimageonagridof512 1024.

The current versionof thisrange cameraop erates at =082 mand atalaser p ower of 150mW.This makes its op eration hazardous to the naked eye. Work has started to enhance the systemto include eye safety and long-range capabilities. Therequirementforeye safety willb e ful lledby op eratingat =154 mandby changingthe silicon-based discrete p ositiondetector byan InGaAsunit. That p osition detectoris describ ed inSection 3.1. With thedata presented inSections3.1and 4.3,the required laserp ower tosaturate thedetector forasurfacewith =1 at adistanceof =10is ab out10mW.At =10m,therequired p ower isintheorderof 1W.Therangecamera willb efurtherenhanced bytheadditionofatime-of- ight(TOF)unitthatwilltakeoverfor 10m. Incidently,a TOF unitprovidesarangeprecisionthat isfairlyconstantwithdistanceas opp osedtoatriangulation-basedcamera where rangeprecision isafunctionof distance(excludingcollected p ower e ects) . Thetransition inprecision o ccurs atab out10m. ThisTOFunitcanb emergedeasilyinthecurrentrangecamera. FromFig.3a,ab eamsplittercanb e p ositioned inthecollectingpathof thelaserb eamjustab ove thep ositiondetector. Andfordistancesover 10m,an avalanchephoto dio dewillpickupthereturnedsignalgeneratedfromapulsedlasersource. Apro ofofconceptunithas b een recentlybuiltto demonstratethetechnique. Atadatarateof10000p ointsp ersecond, therequiredenergy p er pulse forasurfacewith =1at adistance of1mis2 J(including50%loss ofb eamsplitter). At10 m,theenergy isab out 200 J.These p ower orenergy levelsareachievablewithcurrentcommerciallasertechnology. Furthermore, theselevelsare alsowellb elowtheirresp ective MPEswhenthelaserb eamisscanned over ascene (see Section 4.4) .

(18)

a)

Known

Geometrical

Feature

or

Target

Tracking F.O.V. (~ 1°)

130 Measurements / s

(Position / Attitude)

Searching F.O.V. (~ 30°)

1 - 10 s Search Time

Laser

Scanner

High

Precision

Pointing

b)

24 34 :  :   :  7 Conclusion 8 Acknowledgments

Rangecamera: (a)photograph,(b)real-timetrackingofsatellite. Figure11:

Thispap er presents areviewoftheactivitiesindigital3-DimagingintheSWIR(15 mto18 m)at theNational ResearchCouncilofCanada. Theadvantagesoftheauto-synchronizedscanningschemeoverconventionaltriangulation arehighlighted. Inparticular, thelaser sp otp ositiondetection pro cess isreviewedindetailfortwotyp esof p osition detectors that can satisfy di erent system requirements. Design asp ects covering the laser source, optics, and eye-safetyconsiderations are discussed inthe contextof rangeimaging. Three-dimensionalimagingatawavelengthsafe for the humaneye has b een demonstrated. Because of the excellent signal-to-noiseprop erties of theInGaAs lateral e ectphoto dio de,measurementscanb eachievedathighresolutionusingasmallfractionofthemaximump ermissible exp osure (MPE). With the availability of more p owerful laser sources op erating at 1.54 m, one can foresee the developmentof high-sp eed upto video-rate ,high-resolution, andlargevolumeof viewlaserscanners for eye safe rangeimaging. Adiscussiononthecurrentdevelopmentofadual-userangecameraintendedforspaceapplicationsis describ ed. ThiscameracombinesanInGaAsdiscretep ositiondetectorfortriangulationand anavalanchephoto dio de foratime-of- ightunit. Apulsedlasersourceemittingat154 mforeye-safetyconsiderationsisusedbyb othranging devices. Short-(downto0.5m)andlong-range(severalkilometres)measurementswillb ep ossiblewithsuchacamera.

The design of a vision system for eye safe applications is challenging, since it is concerned with all asp ects of evaluation,i.e.,imagequalityanddetection pro cesses. Itinvolvestheunderstandingofallsystemparametersandhow they interactto pro duce results. Rigoroustesting should b e donetodemonstrate that therangecamerasareindeed eye safe. Atthisp ointtheauthorshave notmadeanyapplicationsto thegovernmentagenciesconcerned. This isan essentialstep forvision applicationsthat requireeye safety, ifonewantsto sp eedupthetransition ofvisionsystems fromlab oratorytothespaceenvironmentortothefactory o orwhereeye safetyisaconcern.

The authors wish to thank L.Cournoyer, P. Gariepy, and R. Misner for their technical supp ort alongthe course of these projects. Also acknowledged is P. Amiraultfor thetechnical illustrations. Finally, the authors wish to thank E. Kiddforherhelpinpreparing thetext.

(19)

References 1 23 14(2) 30(11) 959 2067 31(1) 24(6) 665 intensityovershap e.

[1] M.Rioux,\ApplicationsofDigital3-DImaging,"CanadianConf.onElectr.and Comp.Eng.,Ottawa,Sept.4-6, 1990,37.3.1{37.3.10.

[2] P.J.Besl,\RangeImagingSensors,"MachineVisionandApplications, ,127{152(1988).

[3] M.Rioux,\LaserRangeFinderbased onSynchronized Scanners,"Appl.Opt., ,3837{3844(1984).

[4] R.Barib eau, M.RiouxandG.Go din,\ColorRe ectance Mo delingUsing aPolychromaticLaserRange Sensor," IEEETrans.PatternAnal.Mach.Intell., ,263{269(1992).

[5] M.D.Nelson,J.F.JohnsonandT.S.Lomheim,\GeneralNoisePro cess esinHybridInfraredFo calPlaneArrays," Opt.Eng., ,1682{1700(1991).

[6] American NationalStandard for Safe useof Lasers, American NationalStandardsInstitute, NY, ANSI Z136.1-1993.

[7] F.Blais,M.RiouxandJ.-A.Beraldin,\PracticalConsiderationsforaDesignofaHighPrecision3-DLaserScanner System,"Pro c.So c.Photo-Opt.Instrum.Eng. ,225{246(1988).

[8] J.-A. Beraldin, S.F. El-Hakim and L. Cournoyer, \Practical Range Camera Calibration," SPIE Pro ceedings, Videometrics, ,21{31(1993).

[9] J.A. Beraldin, F. Blais, M. Rioux, J. Domey and L. Cournoyer,\Registered Range and Intensity at 10-Mega Samplesp erSecond," Opt.Eng., ,88{94(1992).

[10] G.Bickel, G.Hausler and M. Maul,\Triangulationwith Expanded Range ofDepth," Opt. Eng. ,975-977 (1985).

[11] W. Dremel, G. Hausler and M. Maul,\Triangulationwith Large DynamicalRange," inOptical Techniques for IndustrialInsp ection,P.G.Cielo,Ed.,Pro c.So c.Photo-Opt.Instrum.Eng. ,182-187(1986).

(20)

 

The Infrared Handbook 26(12) 67 30 1735 11(2) 69(1) 50(6) 2020 1683 26(5) 2057 2065 (9) 28(1) 65(2) 30

[13] M.Rioux,G.Bechthold,D.TaylorandM.Duggan,\DesignofaLargeDepthofViewThree-dimensionalCamerafor Rob otVision,"Opt.Eng. ,1245{1250(1987).

[14] J.-A.Beraldin,\DesignofBessel-typ ePre-Ampli ersforLateralE ectPhoto dio des,"Int.J.Electron. ,591{615 (1989).

[15] R.Barib eauandM.Rioux,\CentroidFluctuationsof Sp eckled Targets,"Appl.Opt. ,3752{3755(1991).

[16] G.H.Olsen,\InGaAs llstheNear-IRDetector-arrayVacuum,"LaserFo cus World,MarchA21{A30(1991).

[17] A.M.Joshi,V.S.Ban,S.MasonandM.J.Lange,\512and1024ElementLinearInGaAsDetectorArraysfor Near-Infrared(1 m-3 m)EnvironmentalSensing,"Pro c.So c.Photo-Opt.Instrum.Eng. ,287{295(1992).

[18] F.BlaisandM.Rioux,\Real-TimeNumericalPeakDetector",SignalPro cess ., ,145{155(1986).

[19] G.M.Borsuk,\Photo detectorsforAcousto-OpticSignalPro cessors", Pro c.oftheIEEE, ,100{118(1981).

[20] R.W.Simpson,\NoiseinLarge-ap erture Self-scannedDio deArrays",Rev.Sci. Instrum., ,730{732(1979).

[21] E.R.Fossum andB.Pain,\Infrared ReadoutElectronics for Space ScienceSensors: State oftheArt andFuture directions",InfraredTechnologyXIX, SPIE ,262{285(1993).

[22] A.M.Joshi,G.H.Olsen,V.S.Ban andE.Mykietyn,\Pop cornNoise inLinearInGaAsDetector Arrays",Infrared Fo calPlaneArrayPro ducibilityand RelatedMaterials,SPIE ,200{208(1992).

[23] P. Maigne, J.-A. Beraldin, T.M.Vanderwel, M.Buchanan and D.Landheer,\An InGaAs/InP Position Sensing Photo detector",IEEEJ.ofQuantumElectron., ,820{823,(1990).

[24] F.R.Livingstone,L.King,J.-A.Beraldin andM.Rioux,\DevelopmentofaReal-TimeLaserScanner Systemfor Object Recognition,Insp ection,andRob otControl,"SPIE ,454{461,(1993).

[25] K.S. Ko oijmanand J.L.Horijon,\VideoRateLaser Scanner: Considerations onTriangulationOptics, Detectors andPro cessing Circuits,"SPIE ,251{263,(1993).

[26] J.J. Ewing,\Advanced Solid-State Lasers Challenge Conventional Typ es," Laser Fo cus World, Nov., 105{110 (1993).

[27] J.Hecht,\RareEarths CreateUsefulLong-WavelengthLasers,"LaserFo cusWorld,Nov.,135{142(1993).

[28] M.S.O'Sullivan,J.Chrostowski,E.DesurvireandJ.R.Simpson,\High-p owernarrow-linewidthEr3+-dop ed b er laser,"Opt.Lett.14, ,438{440,(1989).

[29] P.Myslinsky,J.Chrostowski,J.A.KoningsteinandJ.R.Simpson,\HighPowerQ-Switched ErbiumDop ed Fib er Laser",IEEEJ.ofQuantumElectron., ,371{823,(1992).

[30] J.Hecht,\Long-wavelengthDio deLasers areTailoredforFib eroptics,"LaserFo cusWorld,Aug.,79{89(1992).

[31] D.Nitzan,A.E.Brain andR.O.Duda,"TheMeasurementanduse ofRegistered Re ectance andRange Datain Scene Analysis,"Pro c.IEEE, ,206{220(1977).

[32] G.H.Suits,\Natural Sources," Chap.3in by W.L.Wolfe and G.J. Zissis Editors, ERIM, MI,(1989).

[33] M.Rioux,J.-A.Beraldin,M.S.O'SullivanandL.Cournoyer,\Eye-SafeLaserScanner forRangeImaging,"Appl. Opt., ,2219{2223(1991).

[34] F. Blais,J.-A. Beraldin,M.Rioux,R.A.Couvillon,and S.G.MacLean,\Developmentof aReal-Time Tracking Laser Range Scanner forSpace Applications," Workshop on Computer Vision forSpace Applications, Antib es, France,Sept. 22{24,161{171(1993).

Figure

Figure 3 depicts schematically the basic comp onents of two typ es of dual-axis auto-synchronized cameras

Figure 3

depicts schematically the basic comp onents of two typ es of dual-axis auto-synchronized cameras p.5
Figure 7 is used to compute the eect of the dierent input noise sources on the output signals

Figure 7

is used to compute the eect of the dierent input noise sources on the output signals p.13
Figure 10a shows a photograph of the prototyp e that has b een mo died to op erate at 1.5 m

Figure 10a

shows a photograph of the prototyp e that has b een mo died to op erate at 1.5 m p.16
Figure 10b-c shows a recording on a human subject that has b een taken with an incident p ower level of 40 mW ( 900

Figure 10b-c

shows a recording on a human subject that has b een taken with an incident p ower level of 40 mW ( 900 p.16

Références

Updating...