Open House 1998
Remote Sensing and Visualization
of ``Earth from Space''
Welcome to Lamont's new Remote Sensing and Visualization Laboratory!
This facility was established under a grant from the National
Aeronautics and Space Administration (NASA), a grant of computer
equipment from the Intel Corporation, and matching funds contributed
by Lamont and Columbia University. Satellite remote sensing images
collected over the past 25 years provide a way for scientists to
assess environmental changes, such as deforestation in tropical
regions, and relate them to areas of rapid population growth. Our
``RSVLab'', as it is called, contains 21 top-of-the-line color graphics
workstations manufactured by Sun Microsystems, Apple Computer, and
Intel Corporation. Software packages run on these machines provide
an intuitive, menu-driven approach to image processing and analysis
tasks which is extremely helpful to the first-time user.
Today, we illustrate the way in which we can combine satellite remote
sensing imagery with topographical data to make 3-dimensional
stereographic renderings of the Earth's surface. We will use remote
sensing data that comes from the Landsat instrument, which was
flown aboard a series of U.S. satellites since 1972. Landsat, flying
about 500 miles above the Earth's surface, follows a near-polar
orbit to keep pace with the Sun as the Earth revolves beneath it.
Reflected sunlight, both in the visible band that we see, and in the
infra-red band that our eyes cannot see, is recorded by the Landsat
sensor across a ground swath 100 miles wide as the satellite travels
along its orbit. An image is built row-by-row by gathering reflected
sunlight from small 250' x 250' areas positioned side-by-side across
the entire swath, then moving on to the next row of the swath.
For the 3-D stereo example, we take Landsat imagery from part of a
swath crossing the Red Sea coast of Arabia. The small green box locates
the area we will display. We assign the visible green component of
reflected sunlight to a `blue' channel, the visible red component to a
`green' channel, and the reflected infra-red to a 'red' channel. Then we
combine the blue, green and red channels to form a false-color image
of the Landsat data for our area in much the same way as a television
combines blue, green and red components into a color television
picture. To make a 3-D stereo rendering of the Landsat data, we need
a digital representation of the Earth's topographic surface for the
region that includes the Landsat image area. Fortunately, we can use a
280' x 280' grid of topographic height covering southern Arabia
and the Horn of Africa. The most time-consuming part of the process
is to identify distinctive points in both the height grid and the
Landsat imagery that constitute the same locations on the ground in
both data sets. Not surprisingly, these sets of common points are
called 'ground control points' (GCPs). Once we have identified
sufficient GCPs in each data set, we warp or stretch the Landsat data
so that it 'fits' the height data. The Landsat and topography data are
then said to be co-registered, or have the same map projection.
Using the gridded elevation data we make a 3-D perspective view of
the Earth's surface for the small area along the Arabian coast shown
in Figure 1. We first construct the elevated, perspective view of the
terrain using the elevation data. Then we ``drape'' the Landsat imagery
over that surface, rather like placing a carpet over a bumpy floor. So
now we have a perspective view of the area where the false-color
Landsat imagery tells us about the land cover type present in the
area, and the cover type at any particular point can be related to the
elevation at that point using the height data. You're looking in a north-
east direction across the coastal plain towards an elevated, interior plateau.
We have one final step left in making a stereoscopic view of our area.
You'll notice in Figure 2 that even though it is a perspective view of a
3-D surface, it looks kind of `flat'. The modern computer hardware and
software tools in the RSVLab allow us to make 'virtual reality'-type 3-D
rendering of the landscape. First, we make two new images. In one,
each pixel in Figure 1 is shifted to the left by an amount related to the
height value at that pixel. This image is meant for viewing with your
right eye. In the other, each pixel is shifted to the right according to the
height value for viewing with your left eye.
In a moment you'll want to move to the Sun workstation called 'Vega'
to view the stereo 3-D rendering of our area. Perched atop Vega's
tower is a small infra-red signal emitter. Its job is to control the
Crystal Eyes glasses so that the glasses alternately see the left and
right images on the display screen. Your brain does the job of
combining the left and right images into the full 3-D perspective
view of Landsat remote sensing data from the Arabian coast.
So walk over to the computer ``Vega'', don a pair of stereo glasses,
and enjoy the view!