Marc Honikel
WIENER-BASED INTERFEROMETRIC SIGNATURE RECONSTRUCTION
Marc HONIKEL
Federal Institute of Technology, Zurich, Switzerland
Institute for Geodesy and Photogrammetry
marc.honikel@geod.baug.ethz.ch
Working Group II
KEY WORDS: Interferometry, DEM, Simulation, Reconstruction, Data Fusion
ABSTRACT
After almost ten years of regular spaceborne SAR data supply, various microwave remote sensing techniques are
established today. Among them, SAR interferometry (InSAR) and its applications like digital elevation model (DEM)
generation or surface change detection may be the most exciting ones, which the recent shuttle radar topography
mission (SRTM) has proved impressively. SAR interferometry takes advantage from the phase signal properties and is
used to derive the topographic height of an image pixel. The interferometric phase (i.e. the phase difference between
two sensor positions and the target) is a measure for the travel path length difference of the SAR waves proportional to
the target height. Nevertheless, some SAR system inherent limitations corrupt these measurements and leave areas
inaccessible for the interferometric examination, limiting the use of interferometry especially in steep terrain. This paper
addresses the reconstruction of the interferometric signature with a new approach based on the Wiener signal restoration
principle, which distinguishes between more and less noise affected interferometric phase values, thus considering the
partially extreme signal-to-noise (SNR) differences within an interferogram. The interferometric phase is estimated with
help of a synthetic interferogram derived from a stereo-optical DEM, which serves as a complementary data source. In
this way, an estimate of the phase is retrieved even in regions of low signal to noise ratio, which in turn improves the
quality of the InSAR DEM measurement.
The procedure is applied to an interferogram derived from an ERS-1 image pair and a photogrammetric DEM from
SPOT data. All treated phase degradations have been removed enabling precise DEM generation. The resulting DEM
surpassed by far the initial stereo-optical and InSAR DEMs in completeness and accuracy, thus proving the power of
the proposed technique.
1 INTRODUCTION
By mapping about 80% of the landmass of the world interferometrically, the successful shuttle radar topography
mission will help spreading INSAR DEM data into mid-scale databases. In order to derive the height of a point from the
interferometric phase measurements, the phase ambiguities, coming from the 2r measurement interval, must first be
solved with the so-called phase unwrapping, adding the correct multiple of 2x to each phase value. Phase unwrapping
becomes extremely difficult in cases of low SNR, due to signal decorrelation. Although single-pass phase
measurements do not suffer from temporal decorrelation of the phases, which restricts massively the use of InSAR
height measurements from spaceborne repeat-pass systems, the system inherent measurement limitations remain.
InSAR height measurements are restricted to terrain not steeper than the viewing angle of the sensor, leaving
mountainous regions inaccessible for interferometric analysis. In addition to the terrain inclination, the signal
interaction with certain types of vegetation, causing multiple scattering or phase jumps, affect the measurements locally
and act as an additional local noise source.
InSAR height information may serve for the enhancement of existing DEM databases, up to now consisting mainly of
photogrammetrically derived DEMs. Through its generation process, point matching and following interpolation,
stereo-optical DEM quality is less terrain dependent than its SAR counterpart. On the other hand, stereo-optical DEM
generation also suffers from system limitations like cloud occlusions or texture dependency of the matching process,
resulting in measurement holes.
Because of the different data and generation principles, the height measurements of both DEM sources can be assumed
to be uncorrelated. Therefore, when introduced the generation process of their counterpart, the data will enhance the
information content of a scene. The following sections deal with the possibility of overcoming the InSAR limitations by
fusing InSAR with a stereo-optical DEM data, which are used complementarily to the SAR measurements.
The complementarity of both optical and SAR height measurements has already been used for different fusion
approaches for the improvement of DEMs from remote sensing data (Honikel, 1998, 1999). In one case, optical height
148 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000.
me:
add
bot]
The
inte
an
infc
pha
dete
sen:
lim
The
eacl
the
valt
autc
onl:
add
eva.
Cor
thes
dea
layc
whi
fail
2.1
The
ima
(mu
Apr
nois
Cyc!
due