Foreshortening 1
Foreshortening 2
Layover
Figure 1: lllustration of foreshortening and layover. A de-
crease in sensor look angle (from top to bottom) leads to
successively narrower foreshortening areas, and finally to lay-
over.
culate the slopes of the viewed terrain. The application of a
multiple image SAR shape-from-shading algorithm to Mag-
ellan data is reported by [Leberl, 1991]. However, if layover
areas are not recognized as such, the bright layover pixels
caused by superposition of multiple scatterers are misinter-
preted as produced by terrain slopes alone. Therefore, we
believe that the recognition and proper treatment of layover
can be used for a refinement of both stereo and shape-from
shading derived DEMs. An improvement of the DEM in lay-
over areas can be of special interest in planetary sciences,
since steep slopes which lead to layover are often particularly
interesting features for geological and geophysical interpreta-
tions.
In the following we present a concept for the automated de-
tection and exploitation of layover in Magellan SAR imagery,
and describe the first steps of its implementation. In particu-
lar, we address the use of simulation for the generation of test
data, and the development of a matching algorithm specially
suited to SAR foreshortening and layover regions. First tests
on simulated imagery indicate that our matching method can
be utilized for a refinement of stereo-derived DEMs in layover
areas.
284
(a) (b)
Figure 2: A stereo image pair from the Magellan data set,
captured with a look angle of 40 deg (a) and 21 deg (b).
Image size is approximately 17 km x 22 km. The apparent
changes in the central part of the image were classified as
layover by stereoscopy.
(a) (b) (c)
Figure 3: Three views of an area on Venus' surface, acquired
from the left with a look angle of 34 deg (a) and 17.5 deg
(b) during Cycles | and III, repectively, and from the right at
25 deg (c) during Cycle Il. The region marked by an arrow
was classified as layover in (a) and (b) by [Connors, 1994].
2 CONCEPT AND PREVIOUS WORK
Our approach is motivated by a study on the extraction
of height information from Magellan data carried out by
[Connors, 1994]. In that work, discretely dipping surfaces
related to faulting were investigated in the context of geo-
physical applications. The features of interest appear in the
SAR image as bands of increased brightness, which are either
foreshortening or layover areas. The proposed algorithm re
constructs terrain slopes, and at the same time distinguishes
between foreshortening and layover. Its basic idea can be
outlined as follows (see Fig.3): First, corresponding areas
are identified in the Cycle I/1ll same-side stereo pair. Then,
the across-track widths of the two areas of interest are used
along with knowledge about the corresponding sensor look
angle to calculate two possible solutions for the terrain slope.
These two solutions reflect an ambiguity between foreshort-
ening and layover which arises for one of the stereo images
In order to resolve this ambiguity, the opposite-side image aC
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
quire
the «
imag
two ‘
rema
cessi.
of th
denc
The
On t
cessi
heigl
inter
origil
surfa
cura:
areas
appe
two
wher
mall
ing à
layo
widt
socie
thos
stere
revel
to ol
twee
plex
feati
layo
with
scen.
is gi
proje
into
over
corre
merg
prese
teste
be re
The
vide
lustr
impl
newl
the
desig
3.1
In o
ferer
for v
lack
Sive
proc
othe
Up v
