1. INTRODUCTION
In Developing Countries, topographic maps at scales of
1:100,000 and larger are often outdated, if they exist at
all. Topographic mapping and map updating has its
major constraint in the heavy and regular cloud cover
which disables visible and infrared (VIR) sensors to
acquire data from the Earth’s surface. According to a
study by Gastellu-Etchegorry and Ducros-Gambart
(1991), the mean probability of acquiring a remotely-
sensed image of Central Sumatra with less than 30%
cloud cover is only 7%. An operational topographic
mapping programme or monitoring of land cover
changes with optical data is therefore almost impossible.
Since the launch of remote sensing satellites carrying
microwave sensors on board, synthetic aperture radar
(SAR) data are continuously available. Independent from
weather or daylight the active SAR sensors provide
information at any time of the year. Therefore, the
combination of up-to-date radar images with existing
optical data for topographic map updating in Developing
Countries was investigated. The research presented in
this paper used the idea of integrating data from SPOT
and Landsat with ERS-1 and JERS-1 SAR to overcome
the cloud cover problem. Additional benefits of fusing
disparate remote sensing data were the increased
interpretation capabilities and improved reliability of the
results due to the complementary nature of microwave
and optical images. While optical data represent the
reflectance of ground cover in visible and near-infrared,
the radar is very sensitive to the shape, orientation,
roughness and moisture content of the illuminated
ground objects.
2. RESEARCH OBJECTIVES AND
METHODOLOGY
The main objective of the research was to investigate the
geometric aspects of image fusion for map updating in
the humid Tropics. A relevant factor in this respect was
to overcome the cloud cover problem in optical remote
sensing data using microwave imagery. The project
aimed at the production of digital image maps
containing fused image data from different optical and
radar remote sensing satellites. The image data were
geocoded and projected to the Indonesian map projection
system (UTM - ID74). Based on the selected operational
remote sensing satellites LANDSAT, SPOT, ERS-1 and
JERS-1, the resulting image maps were produced at
1:100,000 scale. It is commonly accepted that the scales
between 1:50,000 and 1:250,000 can be achieved with
LANDSAT, SPOT and ERS-1 (Albertz and Tauch,
1994; Dowman et al, 1993; Doyle, 1984; Jacobsen,
1992). The research focused on the geometric aspect of
image fusion evaluating the impact of various parameters
on the fused imagery in terms of geometric accuracy. A
second interest was to find the optimum combination of
satellite data in terms of spatial and spectral resolution.
This included a study on image enhancement
possibilities in the frame of pixel-based image fusion.
The following list comprises parameters which were of
interest to the research:
* Satellite and sensor characteristics;
¢ Geocoding;
* Time;
* Observed ground.
There is a variety of methods to combine different
remote sensing data. The methods identified and selected
during the research period are pixel-based and included
colour transformation techniques (e.g. Red Green Blue
RGB colour composites, Intensity Hue Saturation colour
transformation /HS) as well as statistical/arithmetic
methods (e.g. Principal Component Analysis PCA, band
combinations, ratios). These techniques require a highly
accurate, co-registered or geocoded data set. Especially,
the geocoding of SAR data played an important role
because SAR images are subject to severe geometric
distortions based on the steep viewing geometry of the
sensor. Figure 1 displays a flow chart that describes the
image processing of the remote sensing data to obtain
fused image maps (Pohl and Genderen, 1995).
Data [ RADARSAT || ERS-1 || JERS-1 | | | LANDSAT | spor |
| I bad
|
| System Correction )
SENT Pe |
Pre-Processing :
Geocoding DEM & GCP's
CP's
- Rectification
Polynomial Co-registration Differential
Resampling to common grid
Fusion
Visualization y
Figure 1: Flow chart for fused image map production
After having acquired the image data, they were pre-
processed to remove system induced as well as externally
influenced radiometric and geometric distortions in order
to provide compatible data as input to the image fusion
process. The geocoded data were then introduced to the
actual image fusion process. An evaluation of the
resulting fused imagery considered the topographic map
updating capabilities of the produced fused images. The
best results related to map updating (perceptibility of
topographic features and geometric accuracy) were
printed as image maps including some annotation such
as a coordinate grid and a legend.
3. TEST SITES AND DATA DESCRIPTION
The research relied on the processing of data sets from
two different test sites. The calibration test site is located
656
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
