Terrain seams and normals

[Ludovico]

I'm using LandscapeFetch to get terrain normals, height, etc and orient a node accordingly. Even though the terrain appears smooth, I was surprised this results in bumpy behavior when moved across the terrain. My question is, does the apparent small-scale tiling artifacts affect the normal returned ? For example, in the picture,  it's pretty flat but some triangles in the seam here are 45 degrees off from the vertical, and I am suspecting I get these.

As a follow-up question, if these "seam normals" are indeed returned, Is there a way to query the original heightmap tiling based on the original resolution (ie, one vertex per pixel or so, which would always be smoother)? Thanks again as always.

 

[Ludovico]

I realized the triangulation around tile edges are different, which fooled me into thinking there were little valleys so that was not my issue.

On the same topic (more or less) I increased the Settings -> Texel size from .04 to 1.00 as I have no need for that LoD.

so now I noticed another issue when moving the Texel size up... visual artifacts occur sometimes when viewing along the tile seams. The triangulation seems unusual.

For example, here it looks like there is a huge cut in the hill when there is none as you move closer.




 

Edited December 26, 2020 by Ludovico
typo

[silent]

Can you send us your test project or at least the source textures for heights?

[Ludovico]

@silent I can reproduce by looking straight down most hilly tile seam to about 1 KM.  Another example, this time creating a tunnel artifact...   Nothing like this happens on your terrain projects with a 1.0 texel size? I was hoping the strange wireframe would show you the issue, otherwise, I will try to get a test case at some point.

[Ludovico]

By the way the above screenshots were about 5k X and Y off the origin. I moved in closer to origin and still saw some the extreme fanning along the seams. It's not a showstopper but very noticeable in places. One more here where you can see the stitching strategy and as a result, where it starts to create a visual notch in the hill. .

[silent]

We can say more when we would be able to get a closer look into this test scene. It's hard to tell what is going on here, maybe it's a tessellation issue or some bug in the heights calculation algorithms.