Our research has two major thrusts. The first involves design of instrumentation
to image the earth and development of methods to extract quantitative information
from remotely sensed data. The second involves development of methods to
generate synthetic images of what the earth would look like to airborne
or satellite imaging systems. These synthetic image generation tools include
models of the thermal and radiometric behavior of the earth and the atmosphere
which attempt to simulate all of the important factors that influence the
signal level recorded by sensors operating in the visible through the thermal
infrared portions of the spectrum.
On a recent project, as
part of our synthetic image generation work, we acquired data of an instrumented
scene. Calibrated image data were acquired every hour over a 24-hour period
by sensors operating in several spectral regions. Synthetic representations
of the scene were also produced corresponding to each of the actual images.
While the images matched surprisingly well, it was differences that were
of the most interest. These differences tell us where either our understanding
or modeling of the image-forming phenomena are flawed. This points us to
where we need to study and improve our models.
The modeling efforts are
also valuable in analyzing actual aerial or satellite images. The quantitative
analysis of these images often involve reverse engineering using the same
modeling concepts used in synthetic image generation. In particular, the
removal of atmospheric effects from remotely sensed images often involves
reversing the same procedures used in simulating the atmosphere in synthetic