Page 227 - Digital Analysis of Remotely Sensed Imagery

Image Geometric Rectification 191

1. Pixel height in the DEM
2 O 2. Exterior orientation of the sensor
3. Image coordinate and pixel value
4. Equivalent position on orthoimage

3

A

1

4

FIGURE 5.23 Relationship among a ground point (1), its elevation and
position on the ground and on the remote sensing imagery (3), and the
sensor (2) used in orthorectifi cation.

Any discrepancy between them will translate into errors in the
position of the orthorectified image. Further refinement with the
use of GCPs is accomplished in the same manner as described in
Sec. 5.5.
In comparison to planimetric accuracy, a high vertical accuracy is
more difficult to achieve in orthorectification because it is affected by
more variables. In addition to the regular factors such as the quantity
and quality of GCPs/TPs, and their coordinates inaccuracy, the accu-
racy of image orthorectification is also significantly influenced by the
quality of the DEM and topographic relief. A higher accuracy is
achievable over a gently rolling terrain than over a mountainous
area of a large relief. This accuracy can be as small as RMSE < 2.0 m
(0.9 to 2.0 m) for IKONOS imagery except at the most heterogeneous
site (2.6 m) (Wang and Ellis, 2005), an accuracy level meeting the U.S.
National Map Accuracy Standards for 1:12,000 to 1:4800 maps. How-
ever, the accuracy level (e.g., RMSE at independent check points)
degraded to ±5.1 ~ ±5.7 m for coarse resolution SPOT level 1B stereo
images (Al-Rousan et al., 1997). The residuals in elevation ranged
from ±4.4 to ±7.7 m even when all available GCPs were used in the
absolution orientation.