Frederic Bouvier writes:
> I am playing with SRTM-1 too. I started terrafit with --error=2 and 
> --maxnodes=2000 ( after the patch to terrafit.py ) and I must say 
> it is very impressive to see in action.
> 
> But I found glitches that are quite obvious, and that do not appear
> in the scenery made by Alex Romosan ( certainly with default options,
> before the patch ) : it seems that the buildings are buried in the
> terrain, with small pikes.
> 
> Very visible in downtown SF, between the buildings :
> http://perso.wanadoo.fr/frbouvi/flightsim/fgfs-sf-1arcsec.jpg
> but also visible in Oakland.
> 
> In this one,
> http://perso.wanadoo.fr/frbouvi/flightsim/fgfs-ksfo-1arcsec.jpg
> the maintenance hangar created a hill around it. San Bruno mountains
> are quite impressive though.

I think what you are seeing is the down side to fully automated data
scaning/remote sensing.

As I understand it (which I don't) as the shuttle orbited the earth,
it beamed a radar signal down to the earth and then recorded the
interference patterns in the signal reflection, then did some sort of
fancy math to figure out actual terrain elevation (plus lots of other
crazy stuff to compensate for sensor error.)  It could cover a swath
of about 250 miles if I recall correctly.  They picked an orbit height
that gave them the right coverage for the beam width + a little
overlap.  This orbit also gave them the right spacing between each
path.  The math worked out rather perfectly which is one of the
reasons they flew this mission in the first place.

But when you do this over a city or over building, you can't
necessarily expect to get back the elevation of what the ground would
have been if the buildings weren't there.  Even over forests you can
get a primary reflection off the top of the trees, a secondary
reflection off the underlaying scrub, and maybe a third very faint
reflection off the ground.  I've seen data sets that were nice and
smooth over grassy areas, but got all bumpy and rough over treed
areas.  If you want to calculate flood coverage for instance, you
don't really care about the tops of the trees, you really want the
ground level.

For us too, we would have liked to have had the elevation of the
surface of the ground, but we can't expect that is what we always
get.  Especially since SRTM was fully automated collection and post
processing system.

Other people might actually want the top of the folliage (in MN for
instance you could perhaps use this to calculate which portions of
road get sunlight in the winter and for how long.  Excessively shaded
portions of the roadway might build up icey spots so this could give
you (in advance) a list of potential problem areas to concetrate extra
salt/sand/chemicals.

You will also see other areas around KSFO where there are hills /
spikes that shouldn't be there.

If you want to try to automatically smooth the data, you then
sacrifice quality (kind of like bluring a picture.)

We might need to have the ability to go in and hand edit data but if
we do that we will probably toss accuracy out the window -- maybe --
unless someone has a higher standard they can reference against.

It's a *hard* problem ... but I don't want to be the one to go tell
NASA/NIMA their data sucks :-) because it's the best thing going right
now by a long shot.

Maybe the data would have come out better if the astronauts hadn't
emailed it back to earth as executable attachments using MS Outlook.
Ok, cough, sorry, I should not have included this last paragraph. :-)

Regards,

Curt.
-- 
Curtis Olson   HumanFIRST Program               FlightGear Project
Twin Cities    curt 'at' me.umn.edu             curt 'at' flightgear.org
Minnesota      http://www.menet.umn.edu/~curt   http://www.flightgear.org

_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel

Reply via email to