Sie sind auf Seite 1von 42

i

SyntheticApertureRadar(SAR)RawSignalSimulation

AThesisPresentedtotheFacultyofthe
CaliforniaPolytechnicStateUniversity
SanLuisObispo,California

InPartialFulfillment
OftheRequirementsfortheDegree
MasterofScienceinElectricalEngineering

By
AminShoalehvar
June2012

SupportedbyRaytheonSpaceandAirborneSystemsDivision
ii

2008
AminShoalehvar
ALLRIGHTSRESERVED

iii

CommitteeMembership

Title:SyntheticApertureRadar(SAR)RawSignalSimulation

Author:AminShoalehvar

DateSubmitted:June2012

CommitteeChair:Dr.JohnSaghri

CommitteeMember:Dr.JaneZhang

CommitteeMember:Dr.XiaoHuaYu


iv

Abstract
SyntheticApertureRadar(SAR)RawSignalSimulation
Author:AminShoalehvar
Syntheticapertureradar(SAR)rawsignalsimulationisausefultoolforSARsystem
design,missionplanning,processingalgorithmtesting,andinversionalgorithmdesign.This
thesisexploresaSARrawsignalsimulation.Therawsignalsimulationisthesimulatedreceived
signalbeforeanyprocessingwithexceptionofthedownconverter.Thesimulationplaysa
significantroleinstudiesconcerningnoiseandclutterrejectionandcontributestoward
optimizingSARsystemparameters.
TosimulateSARrawdata,aChirpScaling(CS)methodisused.Thismethod[3]first
stretchestheinputsurfacereflectivityofthetargetintheazimuthandrangedirection
respectively.ThenitderivestherawdatabyinverseequalizingthesignalbasedonCSprinciple.
Thismethodavoidsthetimedomainintegraloperationandimprovesthecomputational
efficiency.Asimulationdiagram,calculationandsystematicprocessareproposedinthisthesis.
Finally,simulationresultsarepresentedtoverifytheaccuracyofcalculationsandtheefficiency
oftheprocess.

Acknowledgements

Iwouldliketothankthepeopleandorganizationsthatmadethispossible:CalPolyUniversity,
RaytheonSpaceandAirborneSystemsandDr.Saghriforhelp,supportguidanceand
sponsorshipofthisproject.Iwouldliketothankmywifeandmyfamilyforstrength,energy
andlovetheyprovideme.Finally,IwouldliketothankDr.JaneZhangandDr.XiaoHuaYufor
volunteeringtheirtimetobepartofthethesiscommittee.

vi

TableofContents
TableofFigures..............................................................................................................................vii
Introduction....................................................................................................................................1
PriorSARProgress...........................................................................................................................5
SARSignal........................................................................................................................................7
SARRawSignalModuleandAnalysisinFrequencyDomain........................................................11
SimulationProgram......................................................................................................................17
SimulationResults.........................................................................................................................19
Conclusion.....................................................................................................................................23
FutureWork..................................................................................................................................25
References....................................................................................................................................26
AppendixA:SARSimulationCodeinMATLAB.............................................................................27
AppendixB:FourierandInverseFourierTransformCodinginMATLABusedinSARRawdata
SimulationProcess........................................................................................................................35
vii

TableofFigures

Figure1:ThesurfaceofVenusasimagedbytheMagellanProbeusingSAR
1
............................2
Figure2:SARGeometry(=LookAngle,r=closestdistancefromPtoflightpath,R=distance
fromthePtotheradar)..................................................................................................................7
Figure3:SimulationdiagramofrawdatabasedonCSprinciple.................................................16
Figure4:Cutsofthephasecurvescomparisoninazimuthdirectionand,averageerrorofthe
phasecurve...................................................................................................................................20
Figure5:Cutsofthephasecurvescomparisoninrangedirection,andtheaverageerrorofthe
phasecurve...................................................................................................................................21
Figure6:(a)SurfaceReflectivityofTarget(CalPolyMap)(b)Amplitudeoftherawdataof
extendedscenes(c)SimulatedSARimageofrawdata................................................................22
Figure7:Surfacereflectivityoftarget,amplitudeofrawdataandfocusedimage.....................24
1

Introduction
SyntheticApertureRadar(SAR)isoneoftheadvancedtechniquesofradarimagingthat
wasdevelopedin1950s,andthetechnologywasreleasedtotheciviliancommunitiesin1970s.
SARisusuallyimplementedbymountingasinglebeamformingantennaonamovingplatform
suchasanaircraftorspacecraft,fromwhichatargetsceneisrepeatedlyilluminatedwith
pulsesofradiowavesatwavelengthsanywherefromameterdowntomillimeters.Themany
echowaveformsreceivedsuccessivelyatthedifferentantennapositionsaredetectedand
storedandthenpostprocessedtogethertoresolveelementsinanimageofthetargetregion.
SARisappliedwidelyinmanyareassuchasmilitary,oceanandagriculture.Software
simulation,whichproducessimulativeechoandimages,isaveryimportantformanypurposes
liketestingdifferentimageformationalgorithms,studyingtheinteractionofelectromagnetic
waveswithascenethatisbeingimaged,testingandvalidatingofdifferentsystemdesign
parametersandeconomicalmethodinresearchofSARsystems.
AtthestartoftheprojectinSeptember2007,theSyntheticApertureRadar(SAR)
simulationfrompreviousMasterstheseswascapableofsimulatingandimagingpointtargets
inatwodimensionalplanewithlimitedmobility.Throughthecourseofthisproject,thefocus
wasonimprovingthecomputationalefficiencyandaccuracyoftheSARsimulationsothatit
couldbeappliedtomorecomplex,timesensitivetwodimensionaltargets.
2

Figure1:ThesurfaceofVenusasimagedbytheMagellanProbeusingSAR
1

Allthesesimulationscanberoughlycategorizedintothreegroups.Thefirstgroupworks
intimedomain(perviousthesis).Itcreatesahighprecisionrawdataset.However,thismethod
haslowcomputationalefficiency.Thesecondgroupconcentratesontherawdatageneration
ofextendedscenesthatoperatesinthetwodimensionalfrequencydomain.Thethirdgroup
cansimulateSARrawdatabyusinginverseimagingalgorithminhybriddomain.Thelimitation
forthismethodistheneedoftherealSARimageasinput,soitlackstheflexibilitytosimulate
SARrawdataoftheartificialtargets.
Thispaperproposestheworkthathasbeendoneinreference[3].Futureworkonthis
ongoingprojectwillincludeanalgorithmtocalculatelineofsightlimitationsofpointtargets.
Anotherfieldistooptimizetheprocessofgeneratingtheradarinformation,sothemore
complexandrealistictargetscanbesimulated.Inaddition,amotioncompensationmethodcan
introducearealarbitrarytrajectorydeviationerrorintotherawdatasimulation.
Thereaderswillfindmostofthetechnicalinformationneededforconcise
understandinganddesigningofhighqualityandhighthroughputSARprocessorsinReference
[1].
1. ThisimageorvideowascataloguedbyJetPropulsionLaboftheUnitedStatesNationalAeronauticsandSpaceAdministration
(NASA)underPhotoID:PIA00104.

Thisbook[1]isdividedintothreeparts.Thefirstpartcontainsinformationrelatedtosignal
processingfundamentals,pulsecompressionoflinearFMsignal,syntheticapertureconcepts,
andSARsignalproperties.ThesecondpartcontainsinformationregardingRangeDoppler
Algorithm(RDA),theChirpScalingAlgorithm(CSA),theOmegaKAlgorithm,theSPECAN
Algorithm,andprocessingScanSARdata.ThethirdpartcontainsDopplerCentroidestimation,
andazimuthFMrateestimation.Thesecondreference[2]talksaboutthebasicprinciples
behindonedimensionalrangeandcrossrangeimagingviaapplicationofradarbandwidthand
syntheticaperturerespectively.Italsodiscussesthespecialroleoftheradarradiationpattern
anditsFourierpropertiesusingthewavefrontreconstructiontheory.Thenitprovidesthe
principlebehindSARimaging,anddiscussessystemmodelingandimagingforsquintspotlight
SARandstripmapSAR.Finally,itdiscussestheSARgeometryandmonopulseSARsystem(aSAR
systemthatusestwoormorereceivingradarstorecordechoeddataduetotransmissionfrom
singleradar).Thereference[4]providesabriefdescriptionofchirpscaling(CS)algorithmandit
presentsimageprocessingusingCSalgorithm.Thereferences[5]and[6]discussageneralized
formulationofextendedchirpscalingalgorithm,whichisapplicableforairandspaceborneSAR
processing.
Inthispaper,thereaderswillgetfamiliarwiththepriorSARprojects.Then,theywill
readabriefsummaryofSARrawsignalformationunderSARSignalchapter.Innextchapter
(SARRawSignalModuleandAnalysisinFrequencyDomain),theywillbeintroducedto
equations,processandsimulationalgorithm.InSimulationProgramchapter,thereaderswill
4

seetutorialfortheMATLABcodeinappendixA.Finally,thereaderswillseethesimulation
resultsandconclusion.

5

PriorSARProgress
Theultimategoalofthisongoingprojectistodevelopapieceofcodewhichusesraw
SARsignalstogenerateimagesoftargetsthatcanbeusedforAutomaticTargetRecognition
(ATR)implementations.However,thisprojectstartedoutasaonedimensionalimplementation
ofradarrangefinding.LynnKendrickdevelopedtheonedimensionalradarrangefinding
simulationin2005asherCalPolySeniorproject.LaterBrianZaharrisdevelopedthefulltwo
dimensionalSARsimulationandrangeDoppleralgorithmin2007asaMastersthesis.Zaharris
simulationarrayedpointtargetsinatwodimensionalplaneofazimuthandrange,withthe
platformtravelinginthesameplanealongtheazimuth.Zaharrissimulationalsomadeuseof
KalmanfilteringoftherawSARsignalportionofthecodetoallowaccurateimagingoflimited
mobilitypointtargetsattheirfinalpositions.PaulMasonusedBrianZaharristwodimensional
rangeDoppleralgorithmandadaptedittotwodimensionalobjectssuchgeometricalshapes
andletterscomposedofarraysofpointtargetswiththeazimuthandrangelocationofeach
pointdefinedininputprofiles.Eachpointtargetrequiresareflectiontobecalculatedduring
eachstageintheflightoverthedurationoftheflight.WiththerealisticSARparametersused,
900reflectionsarecalculatedforeachpointtargetinthetwodimensionalSARsimulationand
duetothecomplexityoftheradarreflectionequationused,withoutoptimizationofthecode
eachpointtargetwouldrequire30secondstocalculateeachreflection.Forlargerimages,such
asMSTARimages,whichareofsize128x128,thesimulationwouldtakeoverfivedaysto
complete.MatthewSchlutzusedZaharrisMATLABcodeofpointtargetSARsimulationto
developmorecomplextwodimensionalandeventuallythreedimensionaltargetSAR
6

simulation.InordertomoveintoATR,movingtargetandmorecomplexthreedimensionalSAR
simulations,amajorflawinthetwodimensionalSARsimulationneededtobesolvedfirst.In
thispaper,anewsimulationmethodinhybriddomainwillbediscussed.

SARSignal
Inthissection,thetheoryofSARrawsignalformationwillbebrieflysummarized.The
theoryofSARsystemshasbeenworkedoutindetailelsewhere,sothepreliminarieswillbe
keptataminimumandjustthebasicswillbepresentedtomakethepaperselfconsistent.
TheSARsystemisconventionalpulsedradar,whichtakesadvantageoftherelative
motionbetweensensorandtargettosynthesizeaverylongantennaandtoachieveahigh
crossrange(azimuth)resolution.Eachechoretainsbothitsamplitudeanditsphase.Itisusual
toadoptthecomplexenveloperepresentationinwhichthereceivedsignaliscomplex,withits
realandimaginarypartsobtainedthroughquadraturedemodulationfromtheincoming
bandpasssignal.

Figure2:SARGeometry(=LookAngle, r=closestdistancefromPtoflightpath,
R=distancefromthePtotheradar)
8

LetusconsidertheradarsensorflyingovertheearthasshowninFig.2withconstant
velocityV
p
atanaltitudeh=R
o
Cos
o
(
o
=LookAngle)alongYdirection(Azimuth).Asingle
scatterer,aslongasstayswithintheantennafootprint,producesaseriesofechoeswitharrival
timesandphasedelaysthatarefunctionsofthesensorpositionwithrespecttotheelementary
scatterer.InSAR,therelativemotionbetweensensorandtargetissupposedknown.SAR
algorithmsestimatethebackscatteringcoefficientofanelementarycellbypickingupfromthe
receiveddatathatsampleswiththerightsequenceoftimedelaysandcorrelatingthemwith
thecorrespondingsequenceofphasedelays.Thechangefromonescantoanotherintheecho
timedelayisknownasrangemigration.Thesequenceofphasedelaysoftheechoesthat
comingfromasinglescatterareknownastargetDopplerhistory.
Asmentionedpreviouslythespacecraft(ortheaircraft)movesataconstantvelocityV
p
and
emitssignalattimegivenby:
p() = rcct |

Ip
| exp|]2n - ]nk
2
] (1)
Whereinfisthecarrierfrequency,T
p
andkarethechirpdurationandrate,respectively.The
expressionofthereceivedsignalafterthedemodulationstepsisasfollow:
b(y
i
, r
i
) = ]JyJr y(y, r) exp|-] I
4n
x
] R - ] I
4n
x
] |
AI
I
cTp
+(i
i
- R)
2
| iect |
r
|
-R
cTo
2
|
2
I
y
|
-y
Y
] (2)
Wherein(y,r)arethe(output)azimuthandrangecoordinates.(y,r)arethecorresponding
coordinatesovertheground.Inaddition,(y,r)istheequivalentbackscatteringcoefficient.
9

isthecarrierwavelength.(.)istheantennagroundilluminationpattern(usually
approximatedtorcct|
y
|
-y

]).Y=
x
L
istheazimuthfootprint.Listheeffectiveazimuthlengthof
thephysicalantenna.
Risadistancefromsensorpositiontothegenericpointofthescene.R
0
isadistancefromthe
lineofflighttothecenterofthescene.fischirpbandwidth.Cisthespeedoflight.T
p
isthe
pulseduration.risc/2timesthetimeelapsedfromthepulsetransmission(Sampling
Coordinate).
rcct |
t
I
| = {
1 i |t| <
I
2
u otbcrwisc
(S.1)
R
2
(n) = R
0
2
+ I
p
2
n
2
(S.2)
Theinstantaneousslantrange,R(),changeswithazimuthtime,,accordingto
equation(3.2),inwhichR()isexpressedasahyperbolicfunctionof.Theequationrepresents
thetargettrajectory,indistanceunits,asafunctionofazimuthtime.Theseparationbetween
rangesamplesisc/(2F
r
),whereF
r
istherangesamplingrate.Thismeansthetrajectorymigrates
throughrangecellsduringtheexposuretimeofthetargetinsignalmemory;hence,thename
RangeCellMigrationorRCMcomesfrom.Thismigrationcomplicatestheprocessing,but
ironically,itisanessentialfeatureofSAR.ThisvariationofslantrangewithtimeimposesanFM
characteristiconthesignalinazimuthdirection.Thehyperbolicformoftheslantrange
equationcanbeexpandedinapowerseries,resultinginalinearRCMcomponent,aquadratic
10

RCMcomponentandhigherorderterms.ThefirstgenerationofsatelliteSARprocessorsused
thepowerseriesexpansionoftherangeequationinthetimedomainandtherangeDoppler
domain.Lateritwasdiscoveredthatthehyperbolicformcouldbekeptinalldomains,thereby
improvingtheprocessingaccuracy.However,thepowerseriesexpansionissometimesuseful
foranalysispurposes.
TherangeDoppleralgorithmwasthefirstalgorithmdevelopedforcivilianSARsatellite
processing.Itisstillthemostwidelyusedalgorithmbecauseofitsfavorabletradeoffbetween
maturity,simplicity,efficiency,andaccuracy.However,undercertainconditions,itstwo
disadvantagescanbecomeapparent.First,ahighcomputingloadisexperiencedwhenalong
kernelisusedtoobtainhighaccuracyintheRangeCellMigrationCorrection(RCMC)operation.
Second,itisnoteasytoincorporatetheazimuthfrequencydependenceofSRC,whichcanlimit
itsaccuracyincertainhighsquintandwideaperturecases.
Achirpisasignalinwhichthefrequencyincreases(upchirp)ordecreases(downchirp)
withtime.Thechirpscalingalgorithmwasdevelopedspecificallytoeliminatetheinterpolator
usedforRCMC.Itisbasedonascalingprinciplewherebyafrequencymodulationisappliedtoa
chirpencodedsignaltoachieveashiftorscalingofthesignal.Usingthischirpscalingprinciple,
therequiredrangevariantRCMCshiftcanbeimplementedbyusingphasemultipliesinsteadof
atimedomaininterpolator.ThealgorithmhastheadditionalbenefitthatSecondaryRange
Compression(SRC)canbemadeinazimuthfrequencydependent.Thisbenefitarisesbecause
thedataareavailableinthetwodimensionalfrequencydomainataconvenientstageinthe
processing.
11

SARRawSignalModuleandAnalysisinFrequencyDomain
Theexpressionofthereceivedsignalafterdemodulationstepsisshowninequation(4).
TheFouriertransform(FT)of(4)alongtherangedirectionisshownin(5).
b(y
i
, r
i
) = ]JyJr y(y, r) exp|-
](4n)
x
R -
](4nk)
c
2
(r
i
- R)
2
| rcct |

|
-R
cT
2
|
2
I
y
|
-y

] (4)
E(y
i
, n) = Jb(y
i
, r
i
) exp(-]nr
i
) Jr'
E(y
i
, n) = JJyJr y(y, r) exp|
-]4nR
z
|
2
(
y
i
- y

) Jexp||-
]4n
c
2
1 K(r
i
- R
)
2
| rcct |
2(r
i
- R)
cIp
| exp|-]nr
i
] Jr
i

b =
4n
c
2
K , r
i
- R = u , Jr
i
= Ju , r
i
= u + R
E(y
i
, n)
= JJyJr y(y, r) exp|
-]4nR
z
|
2
(
y
i
- y

) Jexp|-]bu
2
] rcct |
2u
cIp
| exp|-]nu] exp|-]nR] Ju
E(y
i
, n) = JJyJr y(y, r) exp|-](n
+ |
4n
z
1 R|
2
(
y
i
- y

) Jexp|-]bu
2
] rcct |
2u
cIp
| exp|-]nu] Ju
E(y
i
, n) = rcct |
n
bcIp
| exp|
]n
2
4b
| JJyJr y(y, r) exp|-] |n +
4L
z
1 R|
2
|
y
i
- y

+ (S)
Where:
12

RisdistancefromAtothegenericpointofthescene(R = .r
2
+ (y
i
-y)
2
).(y,r)isscene
reflectivitypattern,includingthephasefactorexp[j(4/)].fischirpbandwidth.Cisspeed
oflight.T
p
ispulsedurationtime.(.)isazimuthilluminationdiagramoftherealantennaover
theground.Yisrealantennaazimuthfootprint( =
xRo
L
).R
o
isthedistancefromthelineof
flighttothecenterofthescene.Listheazimuthdimensionoftherealantenna.risc/2times
thetimeelapsedfromthepulsetransmission.isthevariableinrangefrequencydomain,
b =
4nk
c
2
,andrect[t/T]isstandardrectangularwindowfunction.
TheexpressionoftheFTofequation(5)alongtheazimuthdirectionisshownbelow(6):
E(c, n)
= rcct |
n
bcIp
| exp|
j
2
4b
| Juyui (y, i)
2
|
-i
Y.
i2
-
2
| exp(-jy) expI-ji.
i2
-
2
] (6)
Byusing(6),theintegraloperatorcannotbeavoidedbecausereflectivityofthetargetrawdata
isafunctionofrangedirection.Now,wecanrewritetheequation(6)asfollow:
E(c, n) = rcct |
n
bcIp
| exp|
]n
2
4b
| JJr (c, r)0(c, n, r) (7)
Where:
(c, r) = Jy(y, r) exp(-]cy) Jy (8)
0(c, n, r) =
2
|
-cr
.n
i2
- c
2
| expI-]r.n
i2
-c
2
] (9)
13

n
i
= n +
4n
z

BasedonTaylorseriesexpansion,theexpansionoftherawdatatwodimensionalspectrum
phase(neglectingthree,andhigherorderterms),aroundzero(MaclaurinSeries)isasfollow:
Pbosc (E(c, n)) = = r

n
2
+
8nn
z
+ |
4n
z
1
2
- c
2
-
n
2
4b

(o) +

i
(o)
1!
(n -o) +
"(o)
2!
(n - o)
2

i
=
r(n +
4n
z
)

n
2
+
8nn
z
+ I
4n
z
]
2
- c
2
-
n
2b

" =
r

n
2
+
8nn
z
+ I
4n
z
]
2
- c
2
- r(n +
4n
z
)
(n +
4n
z
)

n
2
+
8nn
z
+ I
4n
z
]
2
- c^2
(n
2
+
8nn
z
+ I
4n
z
]
2
- c
2
)
-
1
2b

I =
4nr
z
+
r

n + |
rz(
2
-1)
8n
3
-
1
4b
+n
2
or o = u
E(c, n) = rcct |
n
bcIp
| E
1
(c, n)
E
1
= JJr (c, r)
2
|
-]crz
4n
1 exp|-] |
4nr
z
1 - ] I
r

] n +] |
zr(1 -
2
)
8n
3
+ |
1
4b
1| n
2
| (1u)

14

Where:
=

1 - |
zc
4n
1
2

Wecangettheanalyticalsolutioninazimuthfrequencydomainandinrangetimedomainby
inverseFouriertransforminginrangedirectionasfollow:
b
1
(c, r
i
) = JJr (c, r)
2
|
-crz
4n
| exp | -
](4nr)
z
- ]nK
s
|
2(r
i
- R
s
(c))
c
|
2
] (11)
Where:
K
s
=
K

| 1 + 2zrK

|
1 -
2
c
2

3
1|

istheeffectiveFMchirprateinrange.R
s
is
R
s
(c) = r (1 + C
s
) (11.1)
equivalentrangemodulationinazimuthfrequencydomain,where:
C
s
(c) =
1

- 1
C
s
iscurvaturefactor,whichdescribestheDopplerfrequencydependentpartofthesignal
trajectory.
ThescalingprinciplethatwasdescribedbyPapoulis[4],wherebyafrequency
modulationisappliedtoachirpencodedsignaltoachieveashiftorscalingofthesignal.Using
15

thischirpscalingprinciple,therequiredrangevariantRCMCshiftcanbeimplemented;using
phasemultipliesinsteadofatimedomaininterpolator.Thealgorithmhastheadditional
benefitthatSRC(SecondaryRangeCompression)canbemadeazimuthfrequencydependent.
Thisbenefitarisesbecausethedataareavailableinthetwodimensionalfrequencydomainata
convenientstageintheprocessing.ByusingChirpScalingprinciple,linearfrequencymodulated
signalmultipliedbycorrelatedfrequencymodulatedsignal(CSfactor),resultisstillafrequency
modulatedsignal,onlythephasecenterandthefrequencymodulationratechanges.After
rangeiscompressedwiththenewfrequencymodulatedrate,displacementoccursatthe
locationofthesignal,whichmakestargetrangecurvatureinfrequencydomainthathasthe
sameshapeindifferentrange.Thisisthepurposeofchirpscalingtoequalizealltherange
migrationtrajectoriestoareferencerange(r
ef
).ByusinginverseCSprinciple,wecanfirst
simulaterawdata,whichhasthesamerangecurvature,andthenwecangetpreciserawdata
byinversedequalizingwithchirpscalingfactor.Accordingly,wecanderivatecurvature
equalizingphasefactorfrom(11)and(11.1)byusingthisidea.
O
1
(c, r
i
, r
c]
) = c
|-]I
4n
c
2
]K
s
C
s
I
|
-
rc]
(1+ C
s
)]
2
]
(12)

b
cs
(c, r
i
) = b
1
(c, r
i
)
1
(c, r
i
, r
c]
)
b
cs
= JJr |(c, r)w
2
|
-crz
. 4n
1
2
(c, r)| . c
{-]
4n
c
2
K
s
(1+C
s
)|
|
-(+C
s

rc]
)]
2
]
(1S)

2
(c, r) = c
|-]
4n
x
-]
4n
c
2
K
s
(1+C
s
)C
s
(-
rc]
)
2
]

16

E
cs
(c, n) =
3
(c, n, r
c]
). FI

|(c, r)w
2
|
-crz
. 4n
1
2
(c, r)| (14)
Wherethe
3
is:

3
(c, n, r
c]
) = c
{]
c
2
n
2
16nK
s
(1+C
s
)
-]nC
s

rc]
]

Thesimulationprogramofthetargetrawdatainextendedscenescanbeextractedfrom
equations(11),(13),and(14).Thesimulationalgorithmisshowninfigure3.

Figure3:SimulationdiagramofrawdatabasedonCSprinciple
17

SimulationProgram
ThesimulationprogramwaswritteninMATLABcodebasedonanalgorithmintroduced
intheprevioussection(figure3).TherawdatasimulationprogramplacedinappendixA.
AppendixAshowstheSARrawdatasimulationinMATLABcoding.Thecodingisbroken
intoafewselfexplanatorysections.Atthebeginningusercansetupplottingintimedomain,
frequencydomain,oreitheronebysettingTimePlotsandSpectralPlotsparameterstoone.In
addition,theusercansettheSARairborneparameterssuchascarrierfrequency,sampling
rate,pulserepetitionfrequency,flightduration,Dopplercentroid,platformvelocity,chirppulse
duration,chirpbandwidth,etcinthissection.(Note:OneofthekeyparametersintheSAR
processingisDopplercentroid(f_dc).TheDopplerEffectisduetotherelativemotionbetween
thepointingangleoftheantennaandthetarget,wherethefrequencyofthereceivedechoes
changescomparedtothetransmittedsignals.Equation15expressestheDopplercentroid,
wheretheistheanglebetweentheantennamainbeamandzeroDopplerplane.Allthe
simulationsinthisthesiswerecalculatedbasedonzeroDopplercentroid.)Then,theprogram
readsthetargetbackscatteringcoefficients,andalignsthetargetareatofitintoSARscanarea.
Fromthispoint,theprogramcalculatesSARrawdatabasedonchirpscalingalgorithmthatwas
explainedintheprevioussection.

dc
=
2 I
p
z
sine (1S)
18

CorrelationbetweenequationsinthispaperandparametersintheMATLABcodeareas
follows:
f_a=x
f_r=
PHI
x
=
x
where:x=1,2,3
gamman=(y,r)
lambda=
w_2=W
2
(.)

19

SimulationResults
Inthissection,thesimulationresultsarepresented.Forconvenience,apointtarget
(3x3)hasbeenusedtomakecomparisonbetweenTimeDomainmethodandFrequency
Domainmethod.ThesystemparametersareshowninTable1(valuesarebasedonreal
application).
Pixel
Parameters
3x3 1171x747
U
n
i
t
s
Pixel
Parameters
3x3 1171x747
U
n
i
t
s
CarrierFrequency 5.330E+09 5.330E+09 Hz NearRange 7.99E+05 7.99E+05 m
SamplingFrequency 1.920E+07 1.920E+07 Hz PlatformVelocity 8.55E+02 8.55E+02 m/s
PRF 1.070E+03 1.174E+03 Hz ChirpPulseDuration 2.16E05 2.16E05 Sec
dur 1.000E+00 1.000E+00 Sec ChirpBandwidth 1.62E+07 1.62E+07 Hz

Table1:SARSystemParameters
Basedontheresultinfigure4andfigure5,thesimulationmethodcangetaveragephaseerror
within60degreesinrangedirectionandwithin40degreesinazimuthdirection,sowecanget
theprecisephaseincomparisontothetimedomain.Thisresultshowshowclosephase
calculationsinfrequencydomainaretothetimedomaincalculations.Thismeansthe
frequencydomaincangetphaseerrornomorethan60degreeinrangedirectionandnomore
than40degreeinazimuthdirection.
ToverifytheefficiencyoftheChirpScalingmethod,withthesystemparametersshown
intable1,therawdataofextendedsceneshasbeensimulatedandshowninfigure6.ittakes
upto8.5864secondstosimulate1171x747pixelrawdata,basedonthechirpscalingprinciple
methodinfigure(3)onIntelCore2DualCPUwith2.2GHzfrequency.Theresultsareshown
intable2.
20

Figure4:Cutsofthephasecurvescomparisoninazimuthdirectionand, average
errorofthephasecurve.

21

Figure5:Cutsofthephasecurvescomparisoninrangedirection, andtheaverage
errorofthephasecurve

Pixels
Methods
3x3 1171x747
TimeDomain 15.06Sec +10Hrs
FrequencyDomain 3.52Sec 4.83Sec

Table2:CalculationEfficiencyResult

22

range
a
z
i
m
u
t
h
SAR Input Reflectivity
200 300 400 500 600 700 800
100
200
300
400
500
600
700
800
900
1000
1100
range
a
z
i
m
u
t
h
Raw data
100 200 300 400 500 600 700 800 900 1000
100
200
300
400
500
600
700
800
900
1000
1100
range
a
z
i
m
u
t
h
Focused Image
200 300 400 500 600 700 800
100
200
300
400
500
600
700
800
900
1000
1100

(a)(b)

(c)
Figure6:(a)SurfaceReflectivityofTarget(CalPolyMap)(b)Amplitudeofthe
rawdataofextendedscenes(c) Simulateu SAR image of iaw uata
23

Conclusion
Wecanseefromsimulationresultsthatthechirpscaling(CS)principlemethodcanget
preciseresultcomparetotimedomainsimulation.Theresultsshowedthephaseerrorcould
notbemorethan60degreeinrangedirectionand40degreeinazimuthdirection.This
methodavoidsthetimedomainintegralandbecauseitprocessesinfrequencydomain,ithas
greatcomputationalefficiency.Toverifytheefficiency,Iranasimulation(seeMATLABcodein
AppendixA)ofextendedsceneswiththesurfacereflectivitythatisshowninfigure7and
systemparametersthatareshownintable3.Thesimulationtookabout2.14secondsto
simulate600x932pixelrawdataonIntelCore2DualCPUwith2.2GHzfrequency.
Inaddition,thismethodcangeneratesimulatedrawdatabasedonanytarget
reflectivitymatrixeswithoutusinginverseSARimagealgorithm.

CarrierFrequency 5.33E+09(Hz) ChirpBandwidth 1.60E+07


DataSamplingRate 1.92E+07(Hz) PlatformVelocity 750(m/s)
FlightDuration 2Second ChirpPulseDuration 2.16E05
Table3:SARSimulationSystemParameters

24

range
a
z
i
m
u
t
h
SAR Input Reflectivity
200 300 400 500 600 700
150
200
250
300
350
400
450
range
a
z
i
m
u
t
h
Focused Image
200 300 400 500 600 700
150
200
250
300
350
400
450
range
a
z
i
m
u
t
h
Raw data
200 300 400 500 600 700 800
100
150
200
250
300
350
400
450
500
550

Figure7:Surfacereflectivityoftarget, amplitudeofrawdataandfocusedimage
25

FutureWork
Thereareseveralareasoffutureworkforcontinuationsonthisproject.Animportant
areaoffutureworkistoaddSARsensortrajectorydeviationssotheprogramcansimulate
morerealistictargetandplatform.Secondareaistoaddtargetmotiontothesimulation,sothe
rawdatacanbegeneratedforATRapplication.Third,extendthecodetoincludemore
algorithms.ItwillbeusefultoincludeotherSARprocessingalgorithmsandrunacomparison
betweentheresultsobtainedfromdifferentalgorithms.Fourth,toincludeknownsatellitedata
formats,oroptionofselectingknowndataformats.
ThisprogramrunsunderMATLAB.However,itisnotverydifficulttoconvertthewhole
programtoanotherenvironmentsuchasC++orJava.Thisisthebestsolutiontogetridof
memorymattersandMATLABlimitationsofarrayandmatrixsize.
26

References
[1]I.G.CummingandF.H.Wong,DigitalProcessingofSyntheticApertureRadarData,
Norwood,MA,2005.
[2]M.Soumekh,SyntheticApertureRadarSignalProcessingwithMATLABAlgorithms,Wilet
Interscience,NewYork,1999.
[3]LuHao,CaoNing,LiuWeiwei,WangFei,HuJurong,EfficientSARRawDataSimulationof
ExtendedScenesUsingChirpScalingPrinciple,IEEE2009,The1stInternationalConferenceon
InformationScienceandEngineering.
[4]A.Papoulis,SystemandTransformswithApplicationsinOptics,McGrawHill,NewYork,
1968.
[5]R.KeithRaney,H.Runge,RichardBamler,IanG.Cumming,andFrankH.Wong,Precision
SARProcessingUsingChirpScaling,IEEETRANSACTIONSONGEOSCIENCEANDREMOTE
SENSING,VOL.32,NO.4,JULY1994.
[6]AlbertoMoreira,Member,ZEEE,JosefMittermayer,andRolfScheiber,ExtendedChirp
ScalingAlgorithmforAirandSpaceborneSARDataProcessinginStripmapandScanSAR
ImagingModes,IEEETRANSACTIONSONGEOSCIENCEANDREMOTESENSING,VOL.34,NO.5,
SEPTEMBER1996.

27

AppendixA:SARSimulationCodeinMATLAB
% =========================================================================
% Aut hor A. Shoal ehvar 12/ 1/ 2009
% The pur pose of t hi s scr i pt i s t o gener at e SAR r aw
% dat a usi ng Chi r p Scal i ng Pr i nci pal . The r aw dat a
% wi l l be saved i n a f i l e on t he comput er har d- dr i ve
% f or f ut ur e use by ot her scr i pt or pr ogr ams.
%
% Last Updat e 4/ 30/ 2012
% =========================================================================
cl ear al l
t i c;
Ti mePl ot s = 0; % To dr aw Ti me Pl ot s, Ti mePl ot s shoul d equal t o 1 ot her wi se 0
Spect Pl ot s = 0; % To Dr aw, Spect Pl ot s shoul d be 1 ot her wi se 0
PLOTS = 1; % To Dr aw pl ot s, PLOTS shoul d be 1 ot her wi se 0

% =========================================================================
% SAR Ai r bor ne Par amet er s
% =========================================================================
c = 3. 00e8; % speed of l i ght i n vacuum
f _c = 5. 33e9; % car r i er f r equency ( Hz)
f _s = 1. 92e7; % dat a sampl i ng r at e ( Hz)
PRF = 0. 300e3; % pul se r epet i t i on f r equency
dur = 2. 0; % f l i ght dur at i on
t _p = 2. 16e- 005; % chi r p pul se dur at i on
B = 1. 60e7; % chi r p bandwi dt h ( Hz)
echoes = PRF* dur ; % number of r adar echoes i n dat a f i l e
near _r ange=7. 99e5; %Near Range ( m)
X0=200; % Hal f Tar get Ar ea Wi dt h ( Tar get i s l ocat ed wi t hi n [ Xc- X0, Xc+X0] )
Xc= near _r ange +X0;
t _near = ( 2* near _r ange) / c; % near r ange f ast t i me ( 2 x r ange)
k_r =- B/ t _p; % Range Chi r p Rat e
dt =1/ ( 2* f _s) ; % Ti me Domai n Sampl i ng I nt er val
Ts=( 2* ( Xc- X0) ) / c; % St ar t t i me of sampl i ng
28

Tf =( 2* ( Xc+X0) ) / c+t _p; % End t i me of sampl i ng


sampl es=2* cei l ( ( . 5* ( Tf - Ts) ) / dt ) ; % number of sampl es per r adar echo( Range)
t au=Ts+( 0: sampl es- 1) * dt ; % Ti me ar r ay f or dat a acqui si t i on ( f ast t i me)

f _dc = 0; % Doppl er cent r oi d ( squi nt angl e)
v = 7. 50e2; % SAR sat el l i t e pl at f or m vel oci t y

r _r ef = ( t au( 1) +sampl es/ 2/ f _s) / 2* c;
al pha = 1. 0;
f _a = - PRF/ 2+f _dc: PRF/ echoes: f _dc+PRF/ 2- PRF/ echoes;
f _r = - f _s/ 2: f _s/ sampl es: f _s/ 2- f _s/ sampl es;
l ambda = c/ f _c;
L_a=1. 0; %Ant enna Lengt h ( m)

% =========================================================================
% Tar get Ar ea
% =========================================================================
t ar get _name=' TG1' ;
[ t ar get 1, map] =i mr ead( t ar get _name, ' gi f ' ) ;
t ar get =doubl e( i nd2gr ay( t ar get 1, map) ) ;
gamman=doubl e( ( t ar get ) ) . / 255;
Naz=si ze( gamman, 1) ; %r ow, X di r ect i on
Nr g=si ze( gamman, 2) ; %col umn, Y di r ect i on
Nt ar get =Naz* Nr g;
cl ear t ar get t ar get 1 map t ar get _name

% Tar get Ar ea Al i gment
dat a=zer os( echoes, sampl es) ;
dat a( cei l ( ( echoes- Naz) / 2) : ( Naz+cei l ( ( echoes- Naz) / 2) - 1) , cei l ( ( sampl es-
Nr g) / 2) : ( Nr g+cei l ( ( sampl es- Nr g) / 2) - 1) ) =gamman;
gamman=dat a;
cl ear dat a

i f ( PLOTS==1)
f i gur e;
29

col or map( gr ay( 256) )


i magesc( gamman) ;
axi s( ' i mage' ) ; axi s( ' xy' )
xl abel ( ' r ange' )
yl abel ( ' azi mut h' )
t i t l e( ' SAR I nput Ref l ect i vi t y' )
end;

% Dat a Fi l e Locat i on
Out put Fi l e_Name=' SAR_RawDat a' ;


% =========================================================================
% Chi r p Scal i ng Al gor i t hm
% Azi mut h FFT
% =========================================================================
gamman = f t x( gamman) ;

i f ( Spect Pl ot s==1)
G=angl e( gamman) ;
xg=max( max( G) ) ; ng=mi n( mi n( G) ) ; cg=255/ ( xg- ng) ;
f i gur e;
col or map( gr ay( 256) )
i mage( 256- cg* ( G- ng) ) ;
axi s( ' i mage' ) ; axi s( ' xy' )
xl abel ( ' r ange' )
yl abel ( ' Doppl er f r equency phase' )
t i t l e( ' r ange si gnal / Doppl er domai n' )
end;

% =========================================================================
% Chi r p Scal i ng
% =========================================================================
D = ( 1 - ( f _a* l ambda/ 2/ v) . ^2) . ^0. 5;
C = 1. / D - 1;
30

R = r _r ef . / D;
C_scl = C + ( 1- al pha) . * ( 1+C) . / al pha;
k_i nv = 1. / k_r - ( 2. * l ambda. * r _r ef . * ( D. ^2- 1) ) . / ( c^2. * D. ^3) ;
k = 1. / k_i nv;


% =========================================================================
% St r et ch sur f ace r ef l ect i vi t y i n azi mut h di r ect i on ( H1* H2) =PHI 2
% =========================================================================
r _0 = t au/ 2* c;
L_s = ( l ambda* r _0) / L_a;
x = k. * C_scl . * ( 1+C) . ^2 . / ( c^2. * ( 1+C_scl ) ) ;
X = x( : ) * ones( 1, sampl es) ;
z = ( r _0- r _r ef ) . ^2;
Z = ones( echoes, 1) * z;
dphi = 4* pi * X. * Z;
H1 = exp( - 1i * dphi ) ;
r _0_scl = r _r ef + ( r _0- r _r ef ) / al pha;
X = ones( echoes, 1) * r _0_scl ;
Z = ( D( : ) - 1) * ones( 1, sampl es) ;
H2 = exp( - 1i * 4* pi / l ambda* X. * Z) ;
w_2=( ( si nc( ( - l ambda* ( f _a) ' * r _0) / ( L_s* 4* pi ) ) ) . ^2) ;
PHI 2= H1 . * H2;
gamman = gamman . * PHI 2;
gamman = di ag( w_2) * gamman;

% =========================================================================
% Range FFT
% =========================================================================
gamman = f t y( gamman) ;

i f ( Spect Pl ot s==1)
G=angl e( gamman) ;
xg=max( max( G) ) ; ng=mi n( mi n( G) ) ; cg=255/ ( xg- ng) ;
f i gur e;
31

col or map( j et ( 256) )


i mage( 256- cg* ( G- ng) ) ;
axi s( ' i mage' ) ; axi s( ' xy' )
xl abel ( ' si gnal f r equency phase' )
yl abel ( ' Doppl er f r equency phase' )
t i t l e( ' 2D spect r al domai n' )
end;

% ========================================================================
% Bul k RCMC, Range Compr essi on: PHI 3
% ========================================================================
x = 1. / ( k. * ( 1+C_scl ) ) ;
X = x( : ) * ones( 1, sampl es) ;
y = f _r . ^2;
Y = ones( echoes, 1) * y;
z = f _r ;
Z = ones( echoes, 1) * z;
A = C( : ) * ones( 1, sampl es) ;
Z = Z . * A;
PHI 3 = exp( 1i * pi * X. * Y) . * exp( - 1i * 4* pi * r _r ef / c. * Z) ;
w_3=si nc( ( f _r * c) / ( 4* pi * k_r * t _p) ) * 1 - 0;
Rect =ones( echoes, sampl es) * di ag( abs( f f t shi f t ( f f t ( f f t shi f t ( w_3. ' ) ) ) . ' ) ) ;
gamman = gamman . * PHI 3;
gamman = gamman . * Rect ;

i f ( Spect Pl ot s==1)
G=angl e( gamman) ;
xg=max( max( G) ) ; ng=mi n( mi n( G) ) ; cg=255/ ( xg- ng) ;
f i gur e;
col or map( j et ( 256) )
i mage( 256- cg* ( G- ng) ) ;
axi s( ' i mage' ) ; axi s( ' xy' )
xl abel ( ' si gnal f r equency phase' )
yl abel ( ' Doppl er f r equency phase' )
t i t l e( ' bul k r cmc and r ange compr essi on' )
32

end;

i f ( Ti mePl ot s==1)
f i gur e;
col or map( j et ( 256) )
mesh( abs( i f t x( i f t y( gamman) ) ) ) ;
xl abel ( ' r ange' )
yl abel ( ' azi mut h' )
zl abel ( ' i nt ensi t y' )
t i t l e( ' af t er H2' )
axi s( ' t i ght ' ) ;
end;

% =========================================================================
% Range I FFT
% ========================================================================
gamman = i f t y( gamman) ;

i f ( Spect Pl ot s==1)
G=angl e( gamman) ;
xg=max( max( G) ) ; ng=mi n( mi n( G) ) ; cg=255/ ( xg- ng) ;
f i gur e;
col or map( j et ( 256) )
i mage( 256- cg* ( G- ng) ) ;
axi s( ' i mage' ) ; axi s( ' xy' )
xl abel ( ' r ange' )
yl abel ( ' Doppl er f r equency phase' )
t i t l e( ' r ange Doppl er Domai n af t er r ange compr essi on' )
end;

% =========================================================================
% Chi r p Scal i ng ( PHI 1)
% =========================================================================
x = k. * C_scl ;
X = x( : ) * ones( 1, sampl es) ;
33

Tau = ones( echoes, 1) * t au;


y = 2. * R. / c;
Y = y( : ) * ones( 1, sampl es) ;
Z = ( Tau- Y) . ^2;
PHI 1 = ( exp( - 1i * pi * X. * Z) ) . ^( - 1) ;

gamman = gamman . * PHI 1;

i f ( Spect Pl ot s==1)
G=angl e( gamman) ;
xg=max( max( G) ) ; ng=mi n( mi n( G) ) ; cg=255/ ( xg- ng) ;
f i gur e;
col or map( j et ( 256) )
i mage( 256- cg* ( G- ng) ) ;
axi s( ' i mage' ) ; axi s( ' xy' )
xl abel ( ' r ange' )
yl abel ( ' Doppl er f r equency phase' )
t i t l e( ' chi r p scal i ng and r ange scal i ng' )
end;

i f ( Ti mePl ot s==1)
f i gur e;
col or map( j et ( 256) )
mesh( abs( i f t x( gamman) ) ) ;
xl abel ( ' r ange' )
yl abel ( ' azi mut h' )
zl abel ( ' i nt ensi t y' )
t i t l e( ' af t er H1' )
axi s( ' t i ght ' ) ;
end;

% =========================================================================
% Azi mut h I FFT
% =========================================================================
gamman = i f t x( gamman) ;
34


% =========================================================================
% Raw Dat a
% =========================================================================
i f ( PLOTS==1)
G=abs( gamman) ;
xg=max( max( G) ) ; ng=mi n( mi n( G) ) ; cg=255/ ( xg- ng) ;
f i gur e;
col or map( gr ay( 256) )
i magesc( ( cg* ( G- ng) ) . / 256) ;
axi s( ' i mage' ) ; axi s( ' xy' )
xl abel ( ' r ange' )
yl abel ( ' azi mut h' )
t i t l e( ' Raw dat a' )
end;
t oc
% End of Raw DAt a Gener at i on ==============================================
dat a = gamman;
PT=t oc;
save( Out put Fi l e_Name, ' dat a' , ' echoes' , ' sampl es' , ' dur ' )

% End =====================================================================

35

AppendixB:FourierandInverseFourierTransformCodinginMATLAB
usedinSARRawdataSimulationProcess
FastFourierTransform(FFT)inXdirection:
%=======================================================
%ForwardFFTw.r.t.thefirstvariable%
%=======================================================
functionfs=ftx(s)
fs=fftshift(fft(fftshift(s)));

FastFourierTransforminYdirection:
%========================================================
%ForwardFFTw.r.t.thesecondvariable%
%========================================================
functionfs=fty(s)
fs=fftshift(fft(fftshift(s.'))).';

InverseFastFourierTransform(IFFT)inXdirection:
%========================================================
%InverseFFTw.r.t.thefirstvariable%
%========================================================
functions=iftx(fs)
s=fftshift(ifft(fftshift(fs)));

InverseFastFourierTransform(IFFT)inYdirection:
%=========================================================
%InverseFFTw.r.t.thesecondvariable%
%=========================================================
functions=ifty(fs)
s=fftshift(ifft(fftshift(fs.'))).';

Das könnte Ihnen auch gefallen