International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B6. Istanbul 2004
Imaging. The Erdas Imagine photogrammetric software
package was then used to automatically generate both a DEM at
the resolution of 10m, and an orthorectified image from the
near-nadir IKONOS image. Figure 2 outlines the procedure.
Image
“Coordinate |
——
/^ Stsupplied
M RPC /
e
in
Yon
Ys
ERDAS &5 |
ficcuracy Assessment
ee
Figure 2. Processing steps and the software tools.
3. TEST FIELD AND THE STEREO PAIR
3.1 Chiengmai Test Field
The city of Chiengmai was selected as a study area for the
evaluation of HRSI metric accuracy for large-scale mapping.
The city is the second largest in the country and is situated
about 700 km north of Bangkok near the foothills of Doi
Suthep. The test field covers an area of about 120 sq. km of the
south-western portion of the city including the airport. The test
area is relatively flat with only about 60m of height variation.
Since all major urban areas in Thailand are also flat, the test
field was considered as well representative for the evaluation.
3.2 The IKONOS Stereo Pair
A stereo pair of IKONOS images were provided to the research
team by GISTDA. Table 1 summarizes relevant information
from the image metadata file.
Details Left Image Right Image
Acquisition Date 17 Mar 2001 17 Mar 2001
Local Acquisition Time | 10:57 10:56
Collection Azimuth 182.0105° 55.733°
Collection Elevation 64.27186° 83.04073°
Sun Angle Azimuth 127.8738° 127.6030°
Sun Angle Elevation 59.16122° 59.00376°
Table 1. Information about the test IKONOS images.
80
3.3 GCPs and Check Points
Initially, a total of 101 well-defined points throughout the
images were selected and their image coordinates measured by
manual pointing. Corresponding ground coordinates were
measured by a combination of rapid static and real-time
kinematic GPS techniques using three Leica SR530 GPS
receivers. The quality of estimated GPS coordinates is about
0.1m or better, though correspondence of the measured feature
point with the image feature point was probably at about the
50cm level. Figure 3 shows an example of these selected points.
Figure 3. Left: A well-defined point on the image. Right:
Corresponding point measured by GPS on the ground.
Figure 4. GCPs and check points.
Points with large standard deviation were discarded to ensure
the integrity of estimated coordinates. There were also points
which could not be determined on the ground with confidence
and these points were discarded. The remaining 81 points were
divided into two groups, 10 points as GCPs and the other 71
points as check points. Figure 4 shows the distribution of these
points.
4. RESULTS FROM BIAS-CORRECTION
Different combinations of GCPs were used in bias-
compensation followed by DEM and orthoimage generation.
The coordinates of checkpoints on the products were measured
and compared with those obtained from GPS. Table 2 lists the
results.