Nuri Al-Nakib
Figure 5: Shaded DEM, Brazil, size of shown area Figure 6: Cadastral map, scale 1 : 5,000
250 x 135 m
3.3 Venezuela
The mapping campaign in Venezuela was carried out in the period of October 1998 until January 1999. 268,000 km? of
the nearly unhabited and inaccessible southern part of the country (states of Amazonas and Bolívar) were mapped with
an image resolution of 5 m x 5 m and a height accuracy of 5 m. The purpose of this project was to provide reference
geoinformation data. The area only was mapped unsufficiently and with less accuracy before, for example in radar
survey projects from 1967 to 1971. The main cartographic products are SAR ortho maps generated from orthorectified
SAR images, and DEM's. In the SAR ortho maps, contour lines with a equidistance of 40 m and a height accuracy of 10
m are embedded. The position accuracy of an image pixel is around 5 m. Due to the dense cloud coverage,
thunderstorms and turbulences, the flight altitude was around 8,000 m. The flown tracks were according to the vast area
up to 130 km long, illuminating a 14 km wide swath. In the project mentioned a very large tropical area was mapped in
order to produce more than 500 SAR ortho mapsheets.
4. GENERATION OF VALUE-ADDED PRODUCTS
The generation of value-added products starts after the SAR data has passed the complete processing and
postprocessing chain. Main inputs for value-added products are both orthorectified SAR images and DEM's. The range
of high-end map products reaches from simple SAR ortho maps (with or without embedded height contour lines, see
figure 7) to cadastral maps (up to a scale of 1 : 5,000, see figure 6), maps with a reduced number of map objects for
planning purposes, to topographic maps in different scales. Other products are three-dimensional views, profiles and
others. Increasingly, digital data is provided according to the specific application for cartographic and, in addition with
further thematic data, for geographic information systems.
Methods of image information extraction are mainly the visual image interpretation, the automatic image classification,
or mostly applied, a mixture of both methods. The visual image interpretation is the most accurate derivation method,
but disadvantages are the long lasting, cost expensive and operator-dependant procedure. On the other side, automatic
image classification algorithms are providing results in a short period of time. As the quality of automatic image
classifications is strongly dependent on the amount and quality of input data, the SAR amplitude image, and
additionally, also coherence image and the correspondent DEM are used. Automatic image classification is more
suitable for small scale product generation (1 : 25,000 and smaller scales, dependant on specific application), as the
cartographic generalization reduces the demand for position accuraccy and map details. A compromise is to use
automatic classification for some map objects like for example water and forest areas only, while objects difficult to
classify (like settlements and the street network) are still interpreted visually. Figure 8 shows a 1 : 50,000 scale map for
planning purposes of the area of Puerto Ayacucho, southern Venezuela. The map was generated by automatic
classification (forest areas, water areas, open areas) as well as by visual interpretation (settlements, road network,
annotation), while the classified areas were coloured and merged with the corresponding orthorectified SAR image.
Figure 9 finally shows a 1 : 50,000 scale topographic map of the same area, which was generated by visual
interpretation only.
The height contour lines generated in the three projects are automatically derived from the corresponding orthorectified
DEM, with possible settings to the equidistance, annotation and other parameters (Schmieder et al., 2000). The
automatically derived height contour lines require later no manual correction or editing. By illuminating with X-band
frequency, the radar signals are reflected by the surface of each object. Therefore, generated DEM's and derived height
contour lines include building and forest heights. By taking attention of the known forest heights and building
recognition, the digital surface model could be corrected to a digital terrain model only representing the ground heights.
22 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000.
Fi
Sc
Gi
Sc
hi
26
Sc
ai
zz