show any drastic withdrawal of glacial borders in FJL and led to
the assumption that the most significant destruction of ice shores
and shelves occured nearly simultaneously in all places and
probably not later than 30 years ago. The impact of present
epeirogenic movements and seismic factors on topographic
changes in the archipelago has already been discussed in
(Kostka, Sharov 1996, b).
3.2 Joint photogrammetric processing of spaceborne optical
and radar images
An original approach called stereophotoradargrammetric has
been offered for further increasing the correctness, detailedness
and consistency of topographic modelling in FJL. This method is
based on merging precision radar imagery with the graphic layer
obtained by stereoplotting from KATE-200 photographs. The
rectified KATE-200 digital image has a ground pixel size of 25
meters, and ERS-I-SAR image data acquired practically
simultaneously over the same area could be registered to the
optical image, since they have similar ground resolutions.
However, direct fusion of these heterogeneous image data did
not provide the desired results. Significant difficulties were
brought about by the different appearance of radar and optical
images having different geometries. Topographic objects above
sea level are displaced towards the subsatellite track in the radar
image and outwards in the photographic image. Therefore, only
layer-by-layer transformation could be applied in order to
correctly perform radar-to-optical image registration, which,
however, would require extremely long computing times due to
the large image size. The lack of reliable information on ERS-1
orbital parameters renders this approach practically impossible.
if only ground-range radar images are available.
Since the transformation of the graphic layer requires much less
computing time, it was decided to combine graphical elements
with the ERS-1-SAR image by transforming the graphic layer
itself. In order to fit the radar image, all planimetric features in
the graphic layer and contour lines were shifted layer-by-layer
towards the ERS-1 subsatellite track. The shift value can be
calculated from the known formula for estimating the layover
displacement in the radar image caused by the influence of relict
D=p- sincos” (122) _sincos (1) | (5)
p pJ |
The meaning of variables D,p,H,Ah is explained in
Figure 5, and equation (5) can be rewritten as follows
D= het 04), (6)
in which the angular measure of layover @ can be
approximately derived as
Ah
a zx —-ctgO: (7)
H 8
The X,Y-components of the shift were calculated in accordance
with the following equations
AX = Ah-ctg0-cos(180 — 8)/ M
and AY = Ah-ctg0 -sin(180 —- B)/ M, (8)
where Ah is the elevation of a contour line with reference to the
current sea level, 0 is the local incidence angle of radar
illumination for the test site, 5 is the local directional angle
(azimuth) of the satellite path and M is the model scale
denominator. The position of the subsatellite track was
preliminarily modelled on the basis of three adjacent radar image
scenes obtained from descending orbit of the ERS-1 carrier, and
the azimuth of satellite flight was estimated as 230? (See Fig. 6).
radar beam
p - slant range
altitude //
; NY
D
"4 le
Fig. 5. Layover distortion of radar image
T
AX X N
E
Flight
-AY direction
— àY
Transposed
graphic symbol À B
se OX P
Corresponding point of Y
radar image
X
Fig. 6. On the calculation of X, Y- shift components and
determination of corrections 0X. 0Y
All shifted contours and planimetric features were superimposed
on the corresponding fragment of an ERS-1 radar image, which
had been appropriately transformed (scaled and oriented)
beforehand to fit the shoreline on the graphic layer. At steep
satellite-faced slopes, the shifted contours sometimes intersect
each other and the shoreline, thus representing the layover
distortion of SAR-images. A cyan mask screened sea-ice floes.
The resultant product was called radar composite (See Attachm.)
Such a product allows for a more reliable interpretation of small
islands, separate rocks, moraines, nunataks, ice shores and ice
divides. The distribution of different glacial zones, including the
belt of superimposed ice and the snow-line position, can be
determined accurately. Besides, inverse radar-to-optical layer-
by-layer transformation has become possible, because a
sufficient number of control points with known elevation are
marked in the radar image. This means that all additional
elements derived from the radar composite can be
cartographically presented in optical image maps. The advantage
of such an approach is that thematic glaciological information,
which can be distinguished in SAR imagery, can be combined
with conventional topographic information to enrich the content
of environmental models.
Investigations on the practical implementation of such an
approach have led to a nontraditional solution to the accuracy
control of stereoplotting and cartographic vectorization. It
simply consists in analyzing the differences (OX, ÓY) between
the positions of the transposed graphical features and those
depicted in the radar image (See Fig.6). Planimetric
206 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
mea
posi
obje
repr
ima
accu
repr
stere
160
140
Fig.
(a) a
The
SNO\
com
intr
cart
solu
sadd
cont
adv:
3.3.
Sate
tool
mon
(Lef
gene
imay
of re
the
diffe
the
