On Jan 31, 2008 10:57 AM, Jon Stockill wrote:
Norman Vine wrote:
> Melchior FRANZ writes:
>> ... and a short article about it: http://www.linux.com/feature/125258
> Just amazing what those gremlins at telascience come up with :-)

Curt needs to get busy with the UAV - not sure how long it'd take to
collect data for our entire default scenery area though ;-)

This is really intriguing ... around the MN areas I checked, the imagery is at least as good as google's.  Certainly there's many areas where the imagery isn't has good, but there is a lot of good stuff there.

It sure would be interesting to develop a variant of our existing scenery system that uses this imagery as the scenery textures rather than our default repeating textures.  It also might be interesting to do a second scenery subsystem based on an entirely different approach with dynamic LOD terrain models and textures that would handle views from ground level up to outer space ... seamlessly.

Let me also point out that a photo real approach is not the answer to everything ... for a lot of reasons that become apparent when you actually try it ... extremely blurry textures in many areas, certain things cooked into the textures like aircraft at an airport.  How do you distinguish water areas versus land areas for gear handling?  How do we place random objects like trees that are appropriate for the surface underneath ... especially when our land use/land cover database is very inaccurate relative to real imagery.

It would actually be nice to support our current approach and a photo-real approach since both have significant advantages and disadvantages ... and perhaps over time we could blend some of the best features of both?


Curtis Olson: http://baron.flightgear.org/~curt/