Page 212 - Digital Analysis of Remotely Sensed Imagery

176 Cha pte r F i v e

FIGURE 5.14 An example of an output image that has been rectifi ed to the
NZMG coordinate system based on a fi rst-order transformation. The edge of
the image appears to be straight. The image’s dimension has increased to
605 rows by 647 columns from 512 rows by 512 columns because of image
rotation. (Copyright CNS, 1994.) See also color insert.

5.6 Issues in Image Georeferencing
Several issues in image georeferencing remain unexplored. For
instance, how does the source of ground control affect accuracy?
What is the relationship between accuracy expressed in absolute
values in meters and that expressed relatively in pixel size? How do
the spatial distribution of GCPs and their ease of identification affect
the accuracy of image georeferencing? In order to answer these
questions, a comparative study was carried out in which a portable
Trimble Geo-Explorer GPS receiver was deployed to repeatedly log
the geographic positions of 20 GCPs in metropolitan Auckland,
New Zealand (Fig. 5.15) and 25 GCPs in a rural area in northwest-
ern China (Fig. 5.16). Working in the autonomous mode with selec-
tive availability, the GPS receiver has a planimetric accuracy up to
100 m, or 2dRMS (approximately 95 percent of the positions are
within the specified value horizontally). The singular horizontal
accuracy can be as high as 2 to 5 m via averaging differentially cor-
rected positions logged within 3 minutes. The urban area of about
2
16 km contains ample road intersections that can serve as GCP can-
didates. Thus, it is relatively easy to select a sufficient number of
widely dispersed quality GCPs. Their coordinates were also read
from a topographic map for the purpose of comparison. The rural
2
scene covering about 94 km encompasses predominantly sand
dunes, and to a lesser extent, cultivated fields. Few distinct features
visible on the satellite imagery can serve as GCP candidates. Scar-
city of these features led to the selection of unstable or inconspicu-
ous water-related features as GCP candidates, such as sharp turning