(NER
NG
tral scanner
od takes an
e corrected
nethod was
inctions. À
ition for the
neters to 2]
be reduced
:d primarily
atively low
iterature for
n (Pope and
F-of-concept
problem of
| the highly
between the
two largely
nethods are
eyer, 1994)
iporally fine
ly improved
el (DEM) to
Yods are the
n (Chen and
s "software
r, they often
ion of GCPs
remes. Iti
; the a priori
e a reference
ser which à
es an initial
Paul Pope
estimate of the position and orientation of the platform as a function of time. The trajectory seed can be derived from
direct measurements of the platform’s position and orientation with time, or this information can be a calculated
estimate of the actual flight path through the use of a few GCPs. The two georeferenced rasters and the input trajectory
constitute a priori information required by the method.
An image-to-image comparison is performed at periodic intervals along the flight path as follows. Scan lines are
synthesized via ray-tracıng for multiple combinations of perturbations in position and orientation about the trajectory
seed. The synthetic scan lines are compared with the actual scan line to determine a set of best matches. The set of
perturbations associated with the best matches are used to correct the trajectory seed. The corrected trajectory is then
used in a separate process to georeference the airborne scanner imagery through parametric modeling. Output from the
method is an irregularly distributed set of planimetric positions for all pixels, from which a geographically registered
image can be formed. This new method is called "parametric modeling with image-to-image matching" (PMIIM).
2. OBJECTIVE
The objective of this research is to describe an initial proof-of-concept of the PMIIM method's ability to georeference
airborne scanner imagery. The approach taken is to synthesize an airborne scanner image through mathematical
modeling of a hypothetical aerial survey. A digital orthophoto is used as the reference image and serves as a model for
the landcover (i.e. provides spectral information). A DEM serves as a model for the terrain (i.e. provides elevation
information). The platform instabilities are known exactly since the trajectory motions are predefined functions of the
exterior orientation parameters. Therefore, the planimetric position of every pixel in the image is also known exactly.
Noise is added to the trajectory to reduce the accuracy of these data, as if the trajectory were being measured by a
hardware system flying onboard the aerial platform. The noisy trajectory is used as the trajectory "seed" to the PMIIM
method. The corrected trajectory output by the PMIIM method and the synthetic airborne image are then used in a
ray-tracing method to georeference the airborne image. The corrected pixel locations are then compared to their known
locations to provide an assessment of the planimetric accuracy of the PMIIM method. Modifications to the trajectory
perturbations are used to investigate the sensitivity of the planimetric accuracy and compute times to these changes.
3. EXPERIMENTS
3.1 Computer Resources
The numerical experiments performed for this research were conducted on two types of machines. The first has a 600
MHz Pentium III CPU with 1,024 Mbyte of RAM running Linux V6.1. The second has a 200 MHz Pentium Pro CPU
with 512 Mbyte of RAM running Linux V5.1. Data were read and written to a network harddrive over a 100 Mbit line.
Three computer codes were used in this research. These codes are written in the Interactive Data Language (IDL) V5.2
(Research Systems, Inc., Boulder, Colorado). The IDL codes, or "procedures" (*.pro files) are called
"create airborne image.pro", "correlate airborne image.pro", and "correct airborne image.pro". The "create" code
synthesizes an airborne scanner image from a reference image, a DEM, a scanner model, a trajectory, and the number of
scan lines requested. Outputs are an airborne scanner image and a point file containing the planimetric location of each
pixel. The "correlate" code implements the PMIIM method. Inputs are a reference image, a DEM, a scanner model, a
trajectory seed, trajectory perturbations, a scan line skip value, and a minimum correlation value. The output is a
corrected trajectory. The "correct" code georeferences an airborne image. Inputs are an airborne image, a DEM, a
scanner model, and a trajectory. Output is a point file containing the planimetric location of each pixel.
3.2 Study Site Description
The study site for this work is an area located south of the town of Cross Plains, Wisconsin. This region is comprised
of highly varying terrain due to a complex arrangement of fine-textured, dendritic erosional patterns. It is known as the
"driflless area" because the southern extent of glaciation during the last ice age terminated just east of this region. The
landcover is a variegated tapestry of tree covered hills and agrarian plots within the flatter valley areas.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 733