118
Specific to radar is a need to process the individual image for
improved stereo-viewing. This consists of gray-tone processing to
make the two images more similar to one another.
2.7 Stereo Model Setup
One needs a capability to select homologue points in both images and
to clear stereo parallax; computation of approximate values for
platform parameters; programs to compute the best orientation; i.e.
flight and imaging parameters to reconcile the two overlapping images
with the ground control points.
2.8 Radar Image Simulation
Geometrically oriented radar product simulation uses a DEM, radar
flight and imaging parameters and a theoretical backscatter curve to
create a SAR image of mountainous terrain.
2.9 Stereo Ortho Image Production
This consists of the creation of a synthetic stereo mate to an ortho
image based on a DEM. Options are needed to create the mate from
the same radar image as of the ortho photo, or from an overlapping
one.
2.10 Radar Image Rectification
A real radar image is matched to a synthetic radar image and then
geometrically rectified. Output is an ortho image, sometimes also
denoted by "geo-coded image".
2.11 Slope Effect Reduction
By considering the slope effect on the radar grey values the radar
image can also be radiometrically rectified. Essentially this consists of
a difference image between a real and DEM-simulated SAR image.
2.12 Mosaicking
Blocks of images are rectified individually to a common geometric
reference, and rectified images fit together to cover several swaths.
The software support can meaningfully exploit automated image
matching capabilities that need to include provisions to emphasize
look-angle-invariant features.
2.13 Height Extraction
Manual methods of stereo image mensuration need to be provided.
This should be supported by digital image correlation. Radar image
correlation needs to account for dissimilarities of the thematic image
contents, speckle and so-called "edge migration".
2.14 Extraction of Pianimetric Features
Essentially this is a manually controlled process based on an
interactive display for either single images or stereo-pairs. Specific
capabilities may i nclude so-called "mono-plotting", image pre
processing for segmentation based on tone, texture and edges.
3. CURRENT STATUS OF DEVELOPMENT
Most of the components listed in section 2 have been developed by us
so that experiences exist regarding their usefulness, applicability and
limitations. However, these developments are experimental so far:
operational capabilities hardly exist at this time for digital SAR images,
with a few exceptions. The following discusses some aspects of the
software components for SAR image analysis.
Little need exists to discuss general image processing and image
archiving/retrieval. Clearly this is widely available in operational
systems. However, DEM processing is already an entirely "exotic"
capability in typical remote sensing data processing environments. If
DEM processing exists, as in commercial products by Intergraph or
International Imaging Systems, it is not integrated with radar relevant
tasks or image simulation.
Limited DEM/SAR processing has been reported by Guindon et al.
(1980), by Naraghi et al. (1983) and in various contributions of our
own (e.g. Domik et al., 1986). No specific effort has been discussed
regarding processing of ground control information. This is taken as
a routine task, except where "smart control" is concerned and control
is searched in images automatically.
Radar image set-up and single image radargrammetric exploitation has
been described in the context of analog film data (Leberl, 1972). It
has been implemented in several experimental settings, most recently
in PASCAL on a PDP 11 in the context of stereo-radar measurements
(Raggam and Leberl, 1984).
Radargrammetric stereomodel set-up only exists for film images.
Autometric (1982), and earlier Norvelle (1972) have created real time
processing capabilities to exploit a radar stereo-model. An equivalent
exploitation system for digital or "soft-copy" images must be based on
stereo-viewing of soft-copy images. Regarding SAR an initial step has
been set by Fullerton et al. (in print).
Radar image simulation is a widely studied topic. Most efforts
concern SAR system simulation, i.e. the simulation of raw SAR
signals. SAR images are created from these with the help of existing
SAR image "correlators" (Smith et al. 1984). This approach differs
from direct simulation of SAR images (So-called SAR-product
simulation) in the manner done by Domik et al. (1986). For
radargrammetric SAR image analysis the image product simulation
should be preferrable to the original simulation.
Image rectification, stereo SAR ortho images and "geo-coding" are
standard procedures of digital photogrammetry. There is no need to
differ significantly from classical techniques when dealing with SAR.
Guindon et al. (1980), Naraghi et al. (1983), Curlander (1986) and
others have developed experimental capabilities; the system that is
currently available to the authors is near-operational and has been
shown to routinely handle large images with 4000 x 8000 pixels.
Radiometric SAR rectification is a concept that has so far not been
used widely: we have a near-operational ability as part of "geo
coding" to also alter the gray-values for effects of terrain slopes
(Domik et al., 1984a). Mosaicking requires that non-similar images
be matched; otherwise it is simply the processing of individual images.
The dissimilarity of overlapping SAR-images is only now becoming a
subject of study as reported by Fullerton et al. (in print).
The actual information extraction of pianimetric and height data from
the SAR images is part of the operation of the stereo plotter and not
necessarily separate from SAR stereo model set-up or SAR mono
plotting. The actual specific tools of course need to be available to
organize the data. This is most easily achieved by interfacing SAR-
specific software with that of classical geographical data input systems
as provided with photogrammelric analytical plotters.
4. DEMONSTRATION STUDIES
Software has been used to analyze radar images from various
missions. The current authors have processed airborne ESA SAR-
580, spaceborne NASA, Seasat, SIR-A and SIR-B as well as other
images. The hardware environment for exploiting analogue/digital
images is VAX based, employing a Gould FD5000 color display
device. Stereo-observation and measurements are yet with analogue
data on a conventional analytical stereo plotter, Kern DSR-11.
We have reported most of our experiments. Seasat exploitation
addressed single image analysis of ice features leading to ice motion
measurements from image series (Leberl et al., 1982). Seasat images
also served in the testing of stereo SAR capabilities (Raggam, 1985).
SIR-A data were the basis for geometric and radiometric rectification
with the help of DEMs (Domik et al., 1984b), and for the extraction
of a DEM from crossing flight lines (Kobrick et al, submitted).
To this date SIR-B has led to the most elaborate radargrammetric
analyses employing multiple incidence angle imagery. The attention
was focussed on stereo-evaluation (Leberl et al., 1986a,b), and on the
creation of secondary image products (Domik et al., 1986).
Other analyzed imagery was from aircraft and was the subject of
stereoscopic viewing as well as measuring. Data sets were examined
from the SAR system of the Jet Propulsion Laboratory (JPL)
(Raggam, 1985). Other data sets, such as from ESA’s SAR-580
campaign, were subjected to single image rectification.
5. CONCLUSIONS
We have operated on an existing series of software components to
support the analysis of SAR images, be they individual or overlapping
coverages. We continue to work to integrate the components into a
viable, portable capability that should be widely available.
The paper segments the range of required analysis tools and discusses
the segments with their purpose, current implementation status by the
authors and general status in the broad international field. It is
argued that ai
Such system <
such as NAS
ERS-1 and fc
digital aircrafi
Acknowledge
Work on the
various agenc
and Jet Propi
more operatk
same parties.
References
Autometric, (
System Radar
for USA Eng;
pages.
Curlander, J.
Generation o;
Spaceborne Ii
Pasadena.
Domik G. (IS
of a Simulatio
Domik, G.,
Radarbildern
Luftbildwesen
Domik, G., I
SIR-B Experi
Products," I
4.
Fullerton K.,
Image Proces
Remote Sensii
Guindon, B.
Models in SA
on SAR I magi
Kobrick, M.,
with Shuttle I
Pasadena, Cal
Leberl, F. (
Imagery," Oi
Photogramme;
Leberl, F., J
Ice Motion N
Geophysical F
Leberl, F., C
"Multiple In
Stereo-Radarg
Sensing.
Leberl, F., (
Stereomappirq
Shasta," IEEI
Naraghi, M.,
Rectification i
Photogr. Eng.
Norvelle, F.F
Eng., XXXVI
Raggam, J., I
Mapping on t
of Photogrami
Raggam, J.
Stereoradargn
152 pages.
Smith, D., I
Simulation G
Research Lab
under Contrac
