w 4
Canada -
an.gc.ca
ontrol points
:quently, the
images was
jact of GCP
methods of
ated to input
oth axes), as
racy) These
High quality
erential GPS
es meet the
| by the user
S.
s des points
'information
s satellitales
t QuickBird.
disponibles,
ns résultats
d’une carte
les données
ient par des
| 1,4 m dans
S”) (0,10 m
les du 1 : 5
cision de la
les aspects
the different
generation,
information
workstation.
tion method
ated. When
. have to be
ters of the
e acquisition
rative least-
“Ps than the
function of
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
The number of GCPs is a function of different conditions: the
method of collection, the sensor type and resolution, the image
spacing, the geometric model, the study site, the physical
environment, GCP definition and accuracy and the final
expected accuracy. If GCPs are determined a priori without any
knowledge of the images to be processed 50% of the points may be
rejected (Toutin, 2004b). If GCPs are determined a posteriori with
knowledge of the images to be processed, the reject factor will be
smaller (20-30%). Consequently, all the aspects of GCP
collection do not have to be considered separately, but as a
whole to avoid too large discrepancies in accuracy of these
different aspects. For example, differential GPS survey should
not be used to process Landsat data in mountainous study site,
nor should road intersections and 1: 50,000 topographic maps to
be used to process QuickBird images if you expect 1-2 m final
accuracy, etc. The weakest aspect in GCP collection, which is
of course different for each study site and image, will thus be
the major source of error in the error propagation and overall
error budget of the bundle adjustment.
In order to address some aspects of GCP collection (definition
and accuracy) with high-resolution satellite data in operational
environments, a collaborative project within Natural Resources
Canada occurred. Scientists at the Centre for Topographic
Information (CTI), the Geodetic Survey Division (GSD), and
CCRS were evaluating the mapping potential of high-resolution
satellite imagery using CCRS 3D multi-sensor physical model
and QuickBird, the highest resolution satellite images available
to the civilian communities in remote sensing/photogrammetry.
2. STUDY SITE AND DATA SET
2.1 Study Site
The study site is the National Capital Region of Canada (45°
20° N, 75° 45’ W): Ottawa, Ontario in the south-east and the
Gatineau Hills, Quebec in the north-west, separated by the
largest half-frozen Ottawa River (East-West) (Figure 1). This
study is mainly a residential environment on both sides of
Ottawa River, and a forest environment in the Hills. The
elevation range is between 50 m in Ottawa to 300 m in the
Gatineau Hills.
2.2 Data Set
To test the CCRS 3D parametric model with QuickBird data,
panchromatic and multispectral imagery products of Ottawa,
were provided as a courtesy of DigitalGlobe™
(http://www digitalglobe.com). The image (16 km by 15 km)
was acquired February 17, 2002 with a low sun elevation angle
of 19°. QuickBird image was provided as Basic imagery
products, which are designed for users having advanced image-
processing capabilities. DigitalGlobe also supplies QuickBird
camera model information with each Basic Imagery product to
permit you to perform photogrammetric processing such as
orthorectification and 3D feature extraction (Robertson, 2003).
This camera model is only useful for the users who do not have
or develop 3D physical model. Basic imagery is the least
processed image product of the DigitalGlobe product suite; only
corrections for radiometric distortions and adjustments for
internal sensor geometry, optical and sensor distortions have
been performed on each scene ordered, and the image
orientation approximately corresponds to a North-South
direction.
Figure 1. QuickBird panchromatic image over the National
Capital Region of Canada (16 km by 15 km; 0.61
pixel spacing).
QuickBird © 2002 and Courtesy DigitalGlobe.
To evaluate the impact of GCP accuracy in the geometric
correction process, four methods of collection were used.
Specifically:
1. Thirty points were collected from 1:50,000 topographic
map. Points are mainly road intersections (Figure 2) with
image pointing accuracy of few pixels (2-3 m). However,
the predominant error comes the map accuracy of around 10
m;
2. Twenty points were collected from 1-m pixel spacing
orthophotographs provided by the Ministere des Ressources
naturelles du Quebec. Points are mainly the same than with
topographic map collection. However, the predominant
error comes also the orthophoto accuracy of 3-5 m;
3. Fifteen points were collected using a hand-held Global
Positioning System (GPS) receiver (WASS enabled). Points
are precise features such as poles (Figure 3) with image
pointing accuracy of one or two pixels (1 m). These poles
were clearly distinguishable due to their long shadows on
snow. However, the predominant error is the GPS accuracy
of 2 to 3 m; and
4. Thirty-eight points were collected, using a differential GPS
(DGPS) receiver in real-time kinematic and post-processing
modes with better than 0.2 m accuracy. Points are mainly
white lines on the ground (Figure 4) with image pointing
accuracy of better than one pixel (0.5 m).
The rationale for the different GCP collection methods was that
the larger number of GCPs would enable error propagation to
be reduced in the computation of the 3D physical model by
using an iterative least-square adjustment method. In fact, the
more accurate the GCPs the fewer GCPs needed for modelling,
and inversely when the accuracy is worse, the number should
be increased depending also of the final expected accuracy
(Savopol et al., 1994).
837