×

You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Title
Fusion of sensor data, knowledge sources and algorithms for extraction and classification of topographic objects
Author
Baltsavias, Emmanuel P.

International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999
30
Reference image: Geocoded Landsat TM, band 2
Input image: SPOT XS, band 1
Coregistered SPOT image
Fig. 13. MBC of SPOT XS to a geocoded Landsat TM image.
5. OVERLAY AND MERGING OF GEOCODED
MULTISENSOR IMAGES
In this section, examples of straightforward combination of
multitemporal and multisensor/multifrequency SAR image data
are presented. It is shown, that just by proper visualisation of
multiple image products the information content of the
individual images can be partly fused and provide an extended
information content from a visual point of view. Figure 14 and
Figure 15 show false colour composites through super
imposition of multisensor SAR images from the ERS, JERS and
Radarsat and of a multitemporal ERS image dataset,
respectively, all of them acquired over an area south of Graz.
Figures 16 and 17 show an example of merging ERS images
acquired from ascending and descending orbits (cf. section 4.2).
In Figure 16 the contribution of the two images is shown in red
and green. In general, the reliable image content of each image
is maintained, whereas the areas which are affected by layover
pixels in one image are replaced through image content from the
other one. As Figure 17 shows, merging provides SAR
backscatter information almost for the entire area, as only some
valleys and mountain ridges of this high-mountainous area do
not have reasonable backscatter values due to common layover
areas (having bright grey values).
6. CONCLUSIONS
The fusion of multitemporal / multisensor remote sensing
images can be efficiently done, only if the data refer to a
common geometry. An overview of methods that may be
seriously considered for geometric coregistration and geocoding
has been given in section 3. As described there, multiple images
may be registered
• either in image geometry with regard to a selected reference
image
• or in map geometry with regard to a selected map projection
or an already geocoded image.
Until now, geocoding of the individual multisensor images to
the geometry of a topographic map is the most usual procedure
to accomplish comparability. However, the experiences from
past applications and the development and implementation of
sensor-specific parametric modelling procedures have shown
that optimum conditions are still sometimes missing. Although
being a straightforward procedure in general, geocoding of
multisensor images involves certain problems and
inconveniences, as follows:
• The ancillary information which is delivered with the image
data is not accurate enough to allow end-to-end geocoding
without interaction and optimisation. More precise image
acquisition parameters would reduce the need of control
points to a minimum and lead to a reduction of interactive
processes.
• Further, the ancillary data (as well as the image data) do not
have standardised and well documented formats. This
problem is reduced, as commercial software to read the
various data formats becomes increasingly available.
Considerable developments related to this aspect are also
made within the RSG software package.
• For geometric registration and geocoding, reference data
like control points, topographic maps, digital elevation
maps etc. are frequently needed. These are not always up-to-
date concerning accuracy and detail of information and
therefore do not always fulfil the demands.