Development of a method to geographically register
airborne scanner imagery through parametric modeling with image-to-image
matching.
Paul Albert Pope - 2001
[UMI Proquest
Full
Citation]
This research developed a modified version of the
direct georeferencing method whereby less accurate and temporally coarse
trajectory information can be used as input, as compared to the highly accurate
and temporally fine trajectory information which is currently required. This
new method uses a unique image-to-image matching algorithm based on the comparison
of synthetic scan lines to real scan lines to determine corrections to the
error-prone input trajectory. The output trajectory can then be used in the
direct georeferencing method to correct the airborne scanner imagery to a
greater planimetric accuracy than that which would result from using the
uncorrected trajectory. This new method is called “parametric modeling
with image-to-image matching”, or PMIIM. Sensitivity experiments based
on synthetic imagery derived from synthetic trajectories showed that the
PMIIM method was able to affect the necessary trajectory corrections. These
experiments also provided guidance on the selection of optimal input parameters
for using this new approach. The PMIIM method was also used to correct real
whiskbroom scanner imagery taken over two study sites exhibiting large differences
in land cover. The PMIIM method was successful when applied to imagery collected
over the site with heterogeneous land cover, but was far less successful
when applied to the imagery collected over the site with more homogeneous
land cover. Additionally, the PMIIM method was not able to correct a highly
variable trajectory. Ignoring corrections to the roll and pitch parameters
of the trajectory drastically reduced the compute time with no loss in planimetric
accuracy. This new method has been shown to be a viable technique for geographically
registering airborne scanner imagery. However, more research is needed to
improve upon the basic approach developed herein.