trilateration net based on airborne geodetic measurements using
radio-range finders. The accuracy of geodetic determinations in
FJL in relation to the continental network was enhanced from
40-50 meters at the beginning to up to 10-15 meters in the
second campaign. This is still not accurate enough to serve as a
reference for precise differential GPS-measurements in the
archipelago. In contrast to any continental area, the datum plane
used in Franz Josef Land is related to the mean level of the
Barents Sea. In FJL, the mean effect of tidal forces does not
exceed 20 to 30 cm, and the sea level recorded at the time of
survey was usually used as a datum plane for practical
topographic works in the archipelago.
The very changeable meteorological conditions, high winds,
heavy clouds and precipitation, especially in the form of snow,
which masks terrestrial details, characterize the study region. On
the average, precipitation occurs every second to third day and
sunshine is observed only during 25% of possible duration (Atlas
of the Arctic, 1985). These unreliable and unfavourable weather
conditions are probably the chief problem for the satellite
monitoring in the High Arctic. Precise photogrammetric
processing of spaceborne image data is further hampered by the
high albedo of glacial landscapes, shadows obscuring details,
and the lack of reliable ground control. Featureless zones and
often almost no visible contrast render image processing
extremely difficult. The accuracies of environmental modelling
via spaceborne imagery and, especially, the accuracy of vertical
measurements and contouring in glacial areas usually do not
meet standard demands. Traditional methods of ground control
surveys, quality control and accuracy tests are ineffective if
applied in the harsh arctic environment.
These are good reasons for attempting to improve the quality of
the retrieval of related topographic parameters from spaceborne
images by means of combining and jointly analyzing image data
obtained in different spectral bands, e.g. visual, infrared and
microwave (multispectral concept), in different seasons and
years (multitemporal concept), at different scales, spatial
resolutions and detail (multisensor or multistage concept).
Experiments in the past have shown the combined use of
spaceborne high-resolution stereophotographs and all-weather
radar imagery to be very expedient for the objective detection,
classification and visual representation of terrestrial changes as
well as for the verification of results of image interpretation in
the High Arctic (Sharov 1997, b).
The experimental data set included high-resolution stereoscopic
photographs obtained by KATE-200 and KFA-1000 cameras
from Russian *Resource-Fl" satellite as well as complex and
precision images taken by the synthetic aperture radar (SAR)
installed on board the European ERS-1/2 satellites. Besides,
the experimental data set included several panoramic
stereoscopic images obtained by the American intelligence
system CORONA (KH-4) in September 1962/64, and 88 aerial
stereophotographs taken over FJL by an AFA-TE100 camera in
August 1953 and 1958. Thus, the multitemporal remote sensing
images available provided, at least, threefold coverage of the
whole territory of the FJL archipelago and sevenfold coverage of
the key-sites. The main parameters of available remote sensing
image data are specified in Table 3. All necessary cartographic
materials and ground-truth data were also available.
Table 3. Description of available remote sensing image data
Camera system Image type Date / time of survey, GMT | Swath, km / resolution, m Quantity
: 07.09.93, 27.08.93 / 15:09 5
KFA-1000 Panchromatic, stereo 28.08.93 / 10:25 80/5—6 10
KATE-200 Multispectral, stereo 25.05.78, 28.08.93 / 10:25 225 / 18.8 — 24 14
CORONA, KH-4(A) Panoramic, stereo 18.09.62, 08.09.64 = 300/4—8 4
AFA-TE100, aerial Panchromatic, stereo 08.05.53, 21.08.53, 23.05.58 5.4/ 0.7 (5 digit.) 88 (27, 52, 9)
ERS-1, SAR Precision 04.09.91, 28.08.93 / 09:20 100/25 —30 5
Complex 03.09.95, 08.10.95 / 09:11 100 / 40 3
ERS-2, SAR Complex 04.09.95, 09.10.95 / 09:11 100 / 40 3
3. PRACTICAL APPLICATION OF IMAGE FUSION
TECHNIQUES TO MONITORING TASKS IN FJL
All image fusion techniques are usually categorized according to
the processing level at which the fusion takes place: pixel,
feature, and decision level. In this paper, we investigate
somewhat different approach called interlaced (or interlayer)
fusion, e.g. the combination of some features derived from one
image with another image. In order to illustrate this hybrid
approach and attendant stratagems for image processing the
following examples were chosen:
e image differencing technique (data fusion at pixel level)
applied to multitemporal space imagery, aerial photographs
and available maps; .
e data fusion at feature level applied to KATE-200
stereophotographs and precision ERS-1-SAR images;
e interferometric analysis of complex SAR images and joint
topographic-glaciological interpretation of interferometric
and stereoscopic models.
3.1 Topographic change detection via multitemporal
photographs
Due to the unique properties of visual perception certain
terrestrial changes can be detected and even measured without
any preprocessing steps by separate visual topographic
interpretation of multitemporal photographs and direct
comparison with available maps. Some drastic topographic
changes in FJL revealed in this way have already been discussed
in earlier publications (Dowdeswell et al. 1994; Kostka, Sharov
1996, a).
In this study, a more accurate technique of photogrammetric
comparison has been applied in order to be able to analyze
smaller changes and execute precise areal measurements. The
analog data set was scanned and digitized so as to avoid any
semantic loss and to ensure that the digital imagery used for joint
analysis is of similar ground resolution, scale and size. For
instance, aerial photographs taken in 1953 at an original scale of
1:30,000 and spaceborne KFA-1000 imagery obtained in 1993 at
1:250,000 scale were digitized with 300 DPI and 2500 DPI,
respectively.
Prior to comparison, stereophotographs were reduced to the
normal case in order to remove the distorting effects of camera
tilt. Image transformation was performed using automatic stereo
correlation procedures and the nearest-neighbor resampling
algorithm (Brandstätter 1993). Control points situated at the
edges of the sea-ice floes were used for the levelling of stereo
models. Basic topographic features including glacial borders,
204 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
coa:
vec
feat
unit
4.8
diff
wer
chai
Sev
are
