Imaging
sitions
6° |
7
= ai
otry and
is being
Tilt Mode
athod for
‚ever, the
RS-1 data
asentative
jes due to
the inherent differences in resolution and
imaging geometry between the two sensors.
Further analysis of the SAR data will include:
1) the merging of two flight lines with
opposite look directions to eliminate radar
shadowing in mountainous terrain; 2)
radiometric rectification to correct for local
incidence angle effects; 3) texture analysis
to emphasize variations between land cover
classes.
5. DISCUSSION AND RESULTS
The preliminary focus was on the SPOT
segment of the project since at the time of
printing, the SAR segment was at a
preliminary stage. Results of the SAR
investigation will be presented in the
symposium session and will be reported in a
later publication.
The SPOT ortho-rectification results are
described in some detail as the availability of
the ortho-rectification software and digital
elevation information were unique in the
authors' experience.
The ortho-correction procedure recently
developed at the Canada Centre for Remote
Sensing (CCRS) (Toutin and Carbonneau,
1992) and transferred to PCI software (PCI,
1993) for commercialization was employed.
Simply described, the technique uses a four
step process: 1) reading the digital &
ephemeral data (satellite orbital and path
information) from tape; 2) collecting ground
control points (including elevation
information); 3) calculating the algorithm; 4)
and correcting the data.
The ortho-rectification package is compatible
with a number of the processing levels
provided by the satellite data vendors
(RADARSAT International Inc. in Canada). In
the first attempt at correcting the SPOT PLA
and MLA data, georeferenced data was used.
Once ortho-corrected, these data yielded
Root Mean Squared errors (RMS errors) of 22
m in the Eastings direction and 6 m in
Northings.
451
The MLA data set were included in the
research effort to allow MacMillan Bloedel
staff to familiarize themselves with digital
classification and land cover discrimination
techniques. The RMS errors achieved with
the MLA data were 17 m in Eastings and 11
m in Northings.
An effort to improve the correction was
sought. Discussions with the author of the
correction algorithms suggested
improvements to the accuracies could be
expected the input data was restricted to the
raw or bulk processed levels.
The authors are confident that a second
attempt at ortho-correction with the more
appropriate input data will increase the
accuracies.
6. SUMMARY - PLANNED FUTURE WORK
The use of SPOT PLA for detecting and
accurately mapping forest harvesting features
has been demonstrated. Ortho-correction of
the SPOT data, using the raw and bulk
products as input, will be attempted. Current
results are encouraging and are expected to
fulfil MacMillan Bloedel’s stringent
requirements.
Work will continue to further improve the
correction and information extraction
techniques. SPOT and SAR mapping results
will be compared with traditional
photogrammetric techniques.
SAR data analysis will include the
development of radiometric and geometric
correction methodologies in addition to the
investigation of texture analysis algorithms to
accentuate variations between land cover
classes.
The radiometric and geometric correction of
the simulated RADARSAT and the ERS-1 data
will serve as a demonstration for the possible
capabilities of RADARSAT data. Although the
results of the mapping accuracies with SAR
data are not directly applicable to the
RADARSAT case, the exercise should
demonstrate the general capabilities of radar