Ron Parker heeft geschreven in bericht <37a85246@news.povray.org>...
.....
>High resolution is likely to be necessary, yes.  I'm not sure about knowing
>the exact location of the camera.  Separation distance, distance from the
>terrain, and some info about lens geometry might be necessary, true, unless
>you were willing to cheat and feed it elevations for the summit and the
>lowlands and let it scale its results to fit.  I think even USGS cheats by
>putting reflective markers at known locations in the area to be photographed,
>so that rectification can take place separate from stereocorrelation.



Imagine a glass plate, on both sides is a painted dot.
Set up the cameras and set focus between both dots, this is the average object
distance (Od). Measure the distance between the two cameras, centere of lens to
centre of lens. This is the StereoBase (Sb). You also need the distance from the
lens to the film plane (Id). Next, measure the parallax between the points in
your pix. Then multiply the parallax by Od^2/Sb*ID. This gives the depth between
the two dots on the glass (thickness of the glass)
Biggest problem for the software is to find the corresponding pixels in the two
pictures. I've done this lots of times by hand. I use a kite plus sterocamera to
do some aerial photography.

http://rst.gsfc.nasa.gov/Start.html
This is a site on remote sensing, I have not explored it yet but chapter 11 is
about stereo stuff.

ingo
--
Met dank aan de muze met het glazen oog.

Post a reply to this message