International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
2.2 Image data
Evaluated images are ‘QuickBird panchromatic basic imagery”
with a radiometric correction and a sensor characteristic
correction. The observation date of images is April 22, 2003
and their ground resolution is 0.78m. The area of each image is
450km? (21.2km*21.2km). The off-nadir angle of the forward
view is 28.7 degree and that of the backward view is 27.2
degree. The overlap is about 9096.
2.3 3D model
The evaluated 3D model is an RPC model by using parameters
distributed with images.
2.4 Image measurement
The 60 check points were selected from obvious objects in
images (e.g.. an intersection of lines at parking lot, Fig. 3).
Then 3 dimensional coordinates are measured on the 3D model
by stereoscoping at a digital photogrammetry software. Each
point is measured twice and calculated the average as the
coordinates A. The standard deviation of the differences
between the two measurements is 0.19m in easting, 0.15m in
northing and 0.23m in vertical.
3
Fig. 3 An example of check points
(Left: forward-view, Right: backward-view)
2.5 GPS survey
The check points were also surveyed by fast-static GPS (Fig. 4)
and coordinates B were obtained. The reference station is one
of permanent GPS stations of the GEONET (GPS Earth
Observation Network of the GSI, Hatanaka et al, 2003). The
observation time of each point is 30 minutes and PDOP of each
observation is less than 6.
% # +
©2004DigitalGlobe.
2.6 Evaluation of horizontal spatial accuracy
The horizontal component of error vectors (coordinates A —
coordinates B) are indicated in Fig. 5.
An apparent westbound shift is observed at all points. The
average of the easting component of the shifts is 13.13 m and
that of the northing component is 2.78m. The standard
deviation of the easting component is 0.73 m and that of the
northing component is 0.93 m.
Z
ct
10m
Fig.5 Horizontal component of error vectors
Since all vectors have the same tendency, we tried to correct
the shifts by fixing one point as a ground control point
(GCP). Fig. 6 shows a correction result in which the center
point of images is used as a GCP. The average of the easting
component of shifts decreased to 0.74m and that of the
northing component decreased to 0.79m.
Z
3km e -
LAU |] ln
Fig.6 Horizontal component of error vectors
(after correction)
Inter
2.71
The
coor
1.42
soutl
Fig.
mode
the c
that f
2.8
In th
evalu
GCP
mapr