Infinity Code Forum — High resolution elevation models

Re: High resolution elevation models

2016-08-31T15:35:14Z

Now that's impressive work. It worked really well, many thanks for the update.

Re: High resolution elevation models

2016-08-31T14:17:02Z

I understand why we have a different numbers, so it doesn't matter.

Re: High resolution elevation models

2016-08-31T14:09:37Z

I know about it.
Actually, yesterday I already implement it (6x6).
3x3 is not enough in most cases.

The new version is already available through the built-in update system.
It contains all what we talked about (and much more), except split mesh.

How did you get 192x192?
I have maximum - 164x164.

Re: High resolution elevation models

2016-08-31T10:58:38Z

Can confirm that using a lower poly collision mesh is a good thing. See attached profile.

This was an experiment with a 3x3 grid, I will need to increase the resolution a bit otherwise the point detection is miles out. But it proves the point and scrolling + zoom still works.

Re: High resolution elevation models

2016-08-30T15:34:32Z

I got it. Many thanks.

Re: High resolution elevation models

2016-08-30T15:30:49Z

On its way. Any problems let me know.

Re: High resolution elevation models

2016-08-30T15:26:29Z

Please send it to support@infinity-code.com.
Thank you.

Re: High resolution elevation models

2016-08-30T15:24:47Z

I've got a 1m res example file I can send you. It is a GeoTIFF and quite large (64Mb 7zipped, ~1Gb raw) Can you dm me an email address and I'll put it in we-transfer for you.

Re: High resolution elevation models

2016-08-30T15:19:13Z

I just looked how to works box collider, and found a way to improve it.
Next update will contain it.

I like the idea to simplify the mesh collider, it will be very good for performance.
I will try to implement it.

I understand. We will use our own data sources.

Re: High resolution elevation models

2016-08-30T14:46:31Z

Indeed. I need to make the mesh collider use a much simpler mesh as most of the time in physics. If I set it to a box collider it uses much less cpu but it doesn't work with the elevation very well (As your handy warning says)

Unfortunately our geo-server is internal to our firewall and has data in that we need to be careful about, so I wont be able to get you access to it. I'll see if I can get a high res model to send you, it will probably have to be a sample product as the full data set has a rather restrictive license, but it should be the correct resolution.

Re: High resolution elevation models

2016-08-30T14:10:02Z

Hmm. 400 ms is a lot.
How do you want to use it when it is so slow at high resolution?

Thank you for the explanation about the vertices. It is very useful.

Can I make requests to your server for testing high resolution elevation?

Re: High resolution elevation models

2016-08-30T13:42:51Z

If your going to add options for this one thing to watch out for is the mesh is not sharing vertices between cells currently, I had to change this otherwise I couldn't get it above about 96x96, the vertex and triangle generation needed modifying so that only one vertex is generated for each point rather than four. I still had to have extra vertices on the edges of the sub meshes (otherwise the uv coords go weird) but it made for less duplication and less vertices in the mesh.

Hopefully that makes sense.

Re: High resolution elevation models

2016-08-30T13:30:44Z

Yeah. I ended up making it detect the size from the provided array when calling SetElevationData

I have to admit the performance is not great with the high res model. (see attached screen shot from the profile. The big step up you can see is once I zoomed in enough to get it to load the elevation model. My work laptop is not a fantastically powerful machine. (4 core I7, 16GB ram, Intel graphics)

Re: High resolution elevation models

2016-08-30T13:13:58Z

Hello.

In the current version there is no good way to do it.
As you have seen 32x32 hardcoded.
We can add some public field (eg int elevationResolution), so you can change the resolution without code modification.

Split mesh is a problem, because (theoretically) it can greatly reduce performance and cause problems.
It requires some research. We'll think about it.

I just am wondering what performance you have with 192x192?
If possible send a screenshot of profiler.

High resolution elevation models

2016-08-30T12:46:25Z

We have a 1 meter resolution elevation model of a reasonable area (205km^2). I have managed to hook up the elevation request interceptor example and load our data from a geo-server using WCS.

By default the resolution of the height map is limited to 32x32 shorts. With some hacking of OnlineMapsTileSetControl.cs I've managed to up this to 192x192 which looks a lot better but I am still running into the limits of 64K vertices in a mesh.

What I have done feels rather hacky. Is there a better way to do this? Also is there a way to split the mesh into more than one mesh so we can get better resolution?