Procedural terrains

rsw

Probably doesn't help much, but here are my thoughts on this...

One thing I notice from your experiments with GE images is that the end mesh is not accurate (in my opinion).

Particularly noticable to me is your first ge image of scotland. The lochs appear to have very steep sides, which may be correct in some instances, but also the lighter coloured area is a mountain (An Teallach I believe) but there does not seem to be much height difference relative to its surroundings as I would expect.

I assume that the mesh is generated by colour of the ge image??

ge/gmaps does seem to have detailed elevation data (google map showing part of the same area of scotland showing terrain with contours), but it doesn't seem like this can be easily exported.

I attempted extracting these contours with raster to vector software, but needed quite a bit of time to tidy up to be of use, although the end result seemed reasonably accurate, especially with the gmaps satellite image added as a texture.

I will keep an eye on how this progresses (particularly if you are trying to create an accurate model)

solo

rsw,

You are correct the hieghts are not correct, I am battling to get good source images, the ones you linked above are much better and I will give that a go tomorrow.
I will keep at it to the point that I either succeed or a better idea comes from it.

Below I used a photo and tried the straight .jpg to displacement method, it is not accurate but still a pretty cool looking result.

Source image

The hidden line mesh 67000 faces.

result.

bcoleman001

@solo said:

Fossa,

I would not say it's an artistic approximation but more of a primary image limitation, see the images I get from GE have shadows and the shadows are a very instrumental in getting the displacement map accurate.

Just curious, but have you tried using the Micro$oft Live Maps? They seem to have less shadows. Just click on the aerial button to get the image. http://maps.live.com/?mkt=en-us Just a thought.

rsw

@solo said:

Below I used a photo and tried the straight .jpg to displacement method, it is not accurate but still a pretty cool looking result.

Definitely a cool result, and very interested in how you are getting from the jpg to the su mesh.

Had another go at extracting the contours from google maps terrain images, with the following result:
Tidying up and assigning heights to each contour is tedious for large areas but not bad for smaller sections. I used the contour maps from http://www.earthtools.org/ which are in metres rather than googles which are in feet. This says that these contours were generated from data, if we could find out how the contours are generated maybe they could be imported straight into su...

pav_3j

two words.

thank you.

pav

rsw

Another go at creating an accurate terrain and came up with this:
terrain quick2.jpg

Created this by using this site to create a 30x30 grid of lat, lon and elev data.

Converted the lat/lon to UTM (meters).

Imported into su with the points cloud ruby.terrain quick.skp

Solo, I thought that using the srtm data with 3DEM to give images like this and used with your method could produce some good results...

Gaieus

Richard,

This site uses GE terrain data, doesn't it? (at least to me it seemed so). What way is it better then than importing terrain from GE?

rsw

@gaieus said:

This site uses GE terrain data, doesn't it? (at least to me it seemed so). What way is it better then than importing terrain from GE?

It grabs the elevation data from usgs seamless elevation data sets, which includes SRTM data (among others) According to wikipedia GE uses srtm data and supplementary dem data to fill gaps. So that site does use very similar data to GE.

As you will see from the skp below there is a difference in the geometry obtained from GE and from this site.terrain compare.skp
The geometry from GE is lower detail - (unless you know how to improve this?)

Anyway I was just experimenting to see what could be done

rsw

@rsw said:

(unless you know how to improve this?)

Forget that, zoom in closer in GE and import into SU, then move along in GE and import into SU again, repeat until you've got the area you want (doesn't give a single terrain object and seems to import image as greyscale but is more accurate).

Can't believe I missed that

Gaieus

Yes, I know this workaround but it will also give you rather clumsy results at the end when you try to stitch them together.

It's actually a great feature in GE that the farther you are the less detailed everything is (otherwise it wouldn't be able to stream) but there should be a way (some option in SU) to import higher resolution terrain data.

Interestingly, we have another discussion about this (and related stuff) in another forum with bjornkn & JClemens.

lapx

Gaieus,
Can you post a test file or link to a data file? I would like to test the cloud ruby.

Gaieus

Lapx,

There is an "exmple.dat" file at Didiers site that you can downoad (and it's also fun at the end)
http://www.crai.archi.fr/RubyLibraryDepot/Ruby/em_geo_page.htm
Go down to cloud.rb and it's on the right side.

lapx

Thanks Gaieus,
But I get thid error message after loading the example "ioc_dat.gif" file.
Any idea what the problem might be?

lapx

Ok, got the dat file to download now. Now it does nothing

lapx

Got it to work. Make sure you have the latest ruby
Now I need to find practical ways to apply it.
Where would one get or create the data files where they could be applicable?

Gaieus

What I useit for is to reconstruct ancient ground surfaces that we explore during archaeological excavations. We set up a co-ordinate system and do the surveys ourselves during the digs. xyz co.ordinates. They are then easy to enter through the script. The more accurate the survey is (i.e. the more point we measure) the more precise the mesh is.

lapx

I guess as I was thinking, a powerful use for 3d would be using it to actually displace difficult to model textures i.e. frieze detail on ancient structures. But this may be overmodeling depending on how close you are to the model.