AMODIFIEDKCCAFORCLUTTERSEPARATIONINAIRBORNEMIMOSAR
YuguanHou
1
,FuqiangZhang
1
,YulongFan
2
1SchoolofElectronicsandInf ormationEngineering,Harb inInstituteofTechnology
92WestDazhiStreet,NanGangDi strict,Harbi n,Chi na
2HeilongjiangProvinceAdmi ni strationofSurvey ing,Mappin gandGeoinformation,Harbin,Chin a
ABSTRACT
Thesignalorclutter separation in airborneMIMO SARby
theinformationofthedirectionofdeparture (DOD) or the
dire ctionofarrival (DOA) is able to be of bene fit to the
suppression of the “unw anted” signals. As the signal
separation performance after using the blocking matrix
algorithm
isgreatlyaff ectedbytheestimatedeviationofthe
dire ction.Moreo ver,thedirectionof the“unwanted”signal
isunkno wnexactly inthepracticalapplication.Toalleviate
this situation, in this paper, a Modified Kernel Canonical
Correlatio n Analysis (MKCCA) approach based on the
blocking matrix approach is proposed. Firstly, the
initial
separation result is obtained by applying the blocking
matrix approach. Then, we use KCCA as a further
separationstep.Theazimuthsearchingresultisadoptedas
the criteria of the separation performance. Simulation
results demonstrate the excellent performance of the
proposedapproach,whichcouldseparatethesignalsor the
clutter
“cleaner”.
Index Terms—MIMO SAR, ICA, blocking matrix
appro ach,MKCCA
1.INTRODUCTION
Due to the rapid development of the phased arr ay
antennas,MIMO radar ha s received considerable attention
inrecenttenyears
[1,2]
.TheapplicationofMIMOtechnique
inSARwasproposed in [36]. However, little work on the
signals or clutter separation in MIMO SAR by the
information of the dir ectionofdeparture (DOD) or the
dire ctionofarrival (DOA) has been reported. Independent
component analysis (ICA) is a very eff icient
blind source
separation method which is able to iso late statistically
independent or nonGaussian component from the
multivariate data. It’s als o a very interesting unsupervised
learningalgo rithms,mainlyforblindsourceseparationand
feature extraction
[7]
. Kernel independent component
analysis (KICA) is an algorithm by means of contrast
functionalrelationshiponreproducingker nelHilbertspace
(RKHS)to completethe ICAprocess, including the kernel
canonical correlation analysis (KCCA) and the kernel
generalized variance(KGV) algorithm
[8]
. In thispaper, we
proposeaModified KernelCanonical Correlation Analysis
(MKCCA) based on the blocking matri x appro ach. This
modif ied approach is able to separate the signals or th e
clutter“cleaner”.
2.DATAMODELOFMIMORADAR
Co nsider an collocated airborne MIMO SAR system
with the assumption the narrow band wavefo rm
[6]
. Let
0
τ
denote the group delay(range) of the target. Also , let
0
1
θ
d
denote the direction ofdeparture (DOD) from the transmit
antenna to the target, and
0
1