Real Aerial/Satellite Imagery & DEMs in Torque
by Alice O · in Game Design and Creative Issues · 06/08/2004 (9:36 pm) · 10 replies
I've got detailed DEMs and high resolution aerial imagery of areas in Australia that I'm playing with to make some virtual environments in Torque. Has anyone used one image to cover the whole area, rather than using various textures with the height/fractal/slope or placement options?
#2
http://holodeck.st.usm.edu/vrcomputing/vrc_t/tutorials/
06/09/2004 (8:07 am)
This is covered on the Desmond Fletcher's website.http://holodeck.st.usm.edu/vrcomputing/vrc_t/tutorials/
#3
Tony, As far as I can tell (and I could be wrong :-)) this site explains about texturing the terrain and using different size texels etc but not how to effectively 'drape' an aerial image over my DEM (hieghfield) data. If I just expand the texel size to be the same as the extents of my heightfield I can do it but it flattens the terrain...
09/19/2004 (9:29 pm)
RE: "This is covered on the Desmond Fletcher's website. http://holodeck.st.usm.edu/vrcomputing/vrc_t/tutorials/"Tony, As far as I can tell (and I could be wrong :-)) this site explains about texturing the terrain and using different size texels etc but not how to effectively 'drape' an aerial image over my DEM (hieghfield) data. If I just expand the texel size to be the same as the extents of my heightfield I can do it but it flattens the terrain...
#5
09/20/2004 (3:05 pm)
Thanks Chris, it's not quiet the same thing, creating hightfields from USGS or other DEM sources is no problem. It's using an aerial image of the area as the texture (rather than painting various textures on a small scale) that I've not been able to do yet. That being said, I've just been informed that trueScape (by DigitalFlux) might do the job... fingers crossed! 
#6
The "Jane's USAF" game was (er, is) a great example of draping aerial/satellite images onto real terrain maps. It faithfully reproduces areas of the southwestern US.
It does have some drawbacks. It can't follow all the twists and turns in the Grand Canyon, and so the Colorado River loops up and down some steep cliffs; that part's pretty funny. In urban areas, you see roads and rooftops; any buildings have to be added in.
I'm not sure what technology Janes used to do the terrain. It was very computation-intensive at the time, like Doom3 is today, and it's probably obsolete anyway.
The U S Army tool, MicroDEM, can drape other data onto terrain, but most of the image data costs money while the height data (DEM, SDTS) is free.
If somebody can figure out how to make draping work inside Torque, I would love it. Right now I'm painting ski runs onto the mountainside by hand. I am surrounded by a relief map (you know, grey terrain with simulated shadows), a USGS contour map, USGS aerial photos and the resort's own trail map.
I wouldn't expect a Torque draper to be as detailed as Jane's. One can always hope, but that's a high-poly game and TGE isn't. Ideally, the draping tool would let one specify textures corresponding to specific patterns or regions on the aerial photo or the topo map.
It's a good project idea for someone.
09/28/2004 (3:43 pm)
@Alice, it looks like you and I are on the same quest. Let me know if you get there first.The "Jane's USAF" game was (er, is) a great example of draping aerial/satellite images onto real terrain maps. It faithfully reproduces areas of the southwestern US.
It does have some drawbacks. It can't follow all the twists and turns in the Grand Canyon, and so the Colorado River loops up and down some steep cliffs; that part's pretty funny. In urban areas, you see roads and rooftops; any buildings have to be added in.
I'm not sure what technology Janes used to do the terrain. It was very computation-intensive at the time, like Doom3 is today, and it's probably obsolete anyway.
The U S Army tool, MicroDEM, can drape other data onto terrain, but most of the image data costs money while the height data (DEM, SDTS) is free.
If somebody can figure out how to make draping work inside Torque, I would love it. Right now I'm painting ski runs onto the mountainside by hand. I am surrounded by a relief map (you know, grey terrain with simulated shadows), a USGS contour map, USGS aerial photos and the resort's own trail map.
I wouldn't expect a Torque draper to be as detailed as Jane's. One can always hope, but that's a high-poly game and TGE isn't. Ideally, the draping tool would let one specify textures corresponding to specific patterns or regions on the aerial photo or the topo map.
It's a good project idea for someone.
#7
When you place an urban ortho pohto in torque (assuming buildings as terrain protuberances), you will find that the top view looks OK in the roofs and top of mountains, but as soon as you navigate through the terrain, you will see the walls of the buildings as stretched plastic wrapping a tube with a very unappealing view. the same happens with cannyons or very steep mountains.
The other factor that makes this difficult, is that is very hard to get true perpendicular ortho-images, so you will always have some kind of perspective distortion when you try to wrap the image over the terrain or the city giving the impression that the edges of the building (or mountain) does not match the terrain irregularities. Also this get affected by the resolution of the terrain and the ortho-photo you are trying to match.
In summary, ortho images are good only to give a nice top view of the terrain, but not to navigate in 3D.
I hope this helps...
William.
09/29/2004 (5:29 am)
I have been working with DEM and GIS from some time and I have been trying to figure out this kind of things as well. From the research I have done, I don't think placing an ortho-photo over Torque is going to give the desired effect.When you place an urban ortho pohto in torque (assuming buildings as terrain protuberances), you will find that the top view looks OK in the roofs and top of mountains, but as soon as you navigate through the terrain, you will see the walls of the buildings as stretched plastic wrapping a tube with a very unappealing view. the same happens with cannyons or very steep mountains.
The other factor that makes this difficult, is that is very hard to get true perpendicular ortho-images, so you will always have some kind of perspective distortion when you try to wrap the image over the terrain or the city giving the impression that the edges of the building (or mountain) does not match the terrain irregularities. Also this get affected by the resolution of the terrain and the ortho-photo you are trying to match.
In summary, ortho images are good only to give a nice top view of the terrain, but not to navigate in 3D.
I hope this helps...
William.
#8
The only problem I foresee with this is other objects in the way (trees, houses, cars, roads, etc). These will have very adverse effects on a heightmap. But, if you find a wide open field I would love to see some screenshots of how this works out for you.
Should work fine, I think. As a player you are subconciously comparing the game-world to real-world. Whether it be something that is feasible in real-world (ie. Half Life) or something based on real-world (ie. Need for Speed Underground). If you use terrain that is directly from the real-world, the comparison will be very easy and your player won't be sitting there thinking, "that don't look right."
10/04/2004 (12:38 pm)
I would convert the images to greyscale and make heightmaps out of them. Heightmaps can be created dynamically with very nice results (esp. if based off of real-world data).The only problem I foresee with this is other objects in the way (trees, houses, cars, roads, etc). These will have very adverse effects on a heightmap. But, if you find a wide open field I would love to see some screenshots of how this works out for you.
Should work fine, I think. As a player you are subconciously comparing the game-world to real-world. Whether it be something that is feasible in real-world (ie. Half Life) or something based on real-world (ie. Need for Speed Underground). If you use terrain that is directly from the real-world, the comparison will be very easy and your player won't be sitting there thinking, "that don't look right."
#9
What's your typical resolution?
The best I've been able to do with freely available stuff is 10 m on the DEM data and 1 or 2 m on the photos. Torque's terrain defaults to 6.5 m, IIRC, although it can go down to 1/4 that -- and developers can define any scale they want, as long as they're willing to scale their game objects acccordingly. The terrain resolution in the aeromap gallery looked like 1 m or better.
Do you (Aeromap) generate your own data (DEM, LIDAR, orthophoto), or do you just work from data collected by USGS or other sources?
As you mentioned, generating hi-res imaged terrain takes a lot of time. It probably takes a lot of RAM, too. It would be interesting to see if one can convert shades and patterns from the orthophotos into textured areas in the Torque terrain faile. We could define a standard set of "earth" textures and use them to somehow convert from the photo image to the textured terrain map - kind of an automated Terrain Painter. Hmm.
10/07/2004 (4:58 am)
Wow, cool stuff. The gallery at www.aeromap.com is fascinating. Are those all rendered images, snapshots from 3D flythroughs?What's your typical resolution?
The best I've been able to do with freely available stuff is 10 m on the DEM data and 1 or 2 m on the photos. Torque's terrain defaults to 6.5 m, IIRC, although it can go down to 1/4 that -- and developers can define any scale they want, as long as they're willing to scale their game objects acccordingly. The terrain resolution in the aeromap gallery looked like 1 m or better.
Do you (Aeromap) generate your own data (DEM, LIDAR, orthophoto), or do you just work from data collected by USGS or other sources?
As you mentioned, generating hi-res imaged terrain takes a lot of time. It probably takes a lot of RAM, too. It would be interesting to see if one can convert shades and patterns from the orthophotos into textured areas in the Torque terrain faile. We could define a standard set of "earth" textures and use them to somehow convert from the photo image to the textured terrain map - kind of an automated Terrain Painter. Hmm.
#10
and would very much like to integrate them into TSE.
I'm willing to work with someone to create a fairly reuseable GDAL
layer for the TSE. I have a fair amount of experience with GDAL and
spatial indexing.
GDAL: http://remotesensing.org/gdal/
10/07/2004 (7:34 am)
I work with DTED-0-1-2 and various high res satellite and LIDAR data,and would very much like to integrate them into TSE.
I'm willing to work with someone to create a fairly reuseable GDAL
layer for the TSE. I have a fair amount of experience with GDAL and
spatial indexing.
GDAL: http://remotesensing.org/gdal/
Associate Kyle Carter