ures
The
the
and
and
gital
ging
ated
tion
ping
data
ntly
the
ates.
will
efits
gital
rcial
s of
of
1 to
rmit
and
1 be
ms.
and
the
1 of
will
may
ture
The
the
hard
[bert
Bill
in P.
y of
s on
ON THE DETECTION AND EXPLOITATION OF LAYOVER IN MAGELLAN SAR IMAGERY
Margrit Gelautz, Fritz Weinbergmair, Franz Leberl
Institute for Computer Graphics
Technical University Graz
Münzgrabenstrafe 11
A-8010 Graz
Austria
e-mail: gelautz@icg.tu-graz.ac.at
Commision IV, Working Group 5
KEY WORDS: Remote Sensing, SAR, Matching, Magellan, Layover
ABSTRACT
In this paper we present a concept for the refinement of Digital Elevation Models (DEMs) derived from Magellan SAR images
of planet Venus. We deal with the automated extraction of height information in foreshortening and layover areas, as well as
the associated recognition of layover. A stereo matching algorithm specially suited to the geometric properties of foreshortening
and layover regions was developed and implemented. First tests carried out on simulated layover maps indicate that the match
points obtained can be utilized for estimating terrain height in SAR layover areas.
KURZFASSUNG
In der vorliegenden Arbeit wird ein Konzept zur Verfeinerung von Digitalen Hóhenmodellen, welche aus SAR-Bildern des
Magellan-Projekts erstellt wurden, vorgestellt. Wir befassen uns mit der automatischen Extraktion von Hóheninformation in
Foreshortening- und Layovergebieten, sowie der damit verbundenen Erkennung von Layover. Ein speziell auf die geometrischen
Eigenheiten von Foreshortening- und Layovergebieten zugeschnittener Matchingalgorithmus wurde entwickelt und implemen-
tiert. Erste Versuche, welche auf simulierten Layover Maps ausgeführt wurden, deuten auf eine Verwendbarkeit der Match-
punkte zur Gelànderekonstruktion in Layovergebieten hin.
1 INTRODUCTION imaging geometry, including the layover phenomenon, can
be found in, e.g., [Leberl, 1990] or [Schreier, 1993]. Fig.1
illustrates how a decrease in sensor look angle leads to a suc-
cessive compression of the foreshortening areas, and finally
to layover. Note the reversed positions of A' and B' in the
ground range projection, when compared to the true posi-
tions A and B in the terrain. Further decrease in sensor look
angle would cause a growing of the layover region. According
to [Kropatsch, 1990], one can distinguish between so-called
active and passive layover areas. Active areas are those which
produce layover, because the local terrain slope exceeds the
sensor look angle. Passive layover areas are only affected by
Magellan SAR images were acquired in three cycles, denoted layover because of an adjacent active layover. (In Fig.1, the
as Cycles I, Il and Ill. In Cycles I and III the radar was looking passive regions are part of the flat areas at the bottom and
to the left, whereas in Cycle I| the imaging configuration was — top of the slope.)
right-looking. This means that for many areas on Venus a
same-side stereo pair as well as a corresponding opposite-side
image are available. Information from same-side stereo im-
ages can be extracted by using conventional stereo matching
techniques, which were originally designed for optical data.
Contrarily, SAR images illuminated from opposite sides ex-
hibit a high degree of geometric and radiometric dissimilar-
ities, which obstruct the joint use of such imagery for the
automated reconstruction of topography.
During NASA's Magellan Mission (1989 - 1994) to planet
Venus, more than 95 96 of the planet's surface was imaged
by the onboard radar sensor, resulting in over 400 Gbytes of
SAR imaging data. One of the major goals of the mission was
the computation of a high-resolution map of the whole planet,
a tool which plays a crucial part in the geophysical analysis
of all planetary processes [Ford, 1992]. Therefore, special
attention needs to be paid to the development of techniques
for extracting height information from SAR imagery which
are suited to this particular data set.
In Magellan SAR data a considerable amount of layover can
be found, due to the steep look angle used in Cycles Il and
I| (between 11 deg and 25 deg). Examples of layover in
Magellan images are given in Figs.2 and 3. The layover area
shown in Fig.2 is located on Venus at about 8 deg S, 74 deg
E. The layover was recognized by stereoscopy, since due to
geometric and radiometric perturbations layover areas do not
fuse properly in stereoscopic vision. Fig.3 shows a 26 km x
43 km section of the Venusian surface at 29.5 deg S, 142.5
Radar layover is a special problem that arises when dealing deg E. The technique used to identify the layover regions in
with SAR imagery of mountainous terrain, with slopes steeper (a) and (b) was developed by [Connors, 1994], and will be
than the off-nadir look angle of the SAR sensor. For these briefly described in the next section.
slopes, the top of the mountain is closer to the sensor than the
bottom. Since radar is a range measuring device, this con- When automated matching methods are applied to stereo
figuration leads to particular geometric distortions, which are images, the geometric and radiometric differences associated
denoted as layover. Due to the superposition of multiple scat- ^ with layover result in inaccurate or missing match points,
terers from different parts of the terrain, layover areas appear which lead to errors in the derived Digital Elevation Model
in the SAR image as bright regions with the original geometric (DEM). Similar problems arise when employing shape-from-
order being reversed. A more detailed discussion of the SAR shading techniques, which use the pixel gray values to cal-
283
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
