Masanori Kakimoto (kaki++at++kaki.nsg.sgi.com)
Tue, 19 Mar 1996 15:59:38 +0900
You are right. That's why I said "more detailed LOD tiles are in most
^^^^^^^
cases closer to the viewer," NOT "always."
^^^^^
It is easy to show such example geometry on which the scheme does
not work. But natural terrain is usually not so nasty.
I tried the trick for some real elevation data of some district full
of high, steep mountains where the 1998 Winter Olympic Games will be
held. The result is "it is difficult to see the background through
those gaps" as I said before. It is possible to find the gaps when I
look almost vertically down onto the steep area. I felt the frequency
is small enough to be acceptable.
If you really need to avoid the phenomenon, limiting the pitch of
viewing direction out of e.g. (-85, -95) while you fly will make it
more difficult to see the hole. Although some military flight
simulator trainees may complain about this limitation.
>
> But, I hear you say, A is always further away than B - ergo it's always at
> lower detail than B - so no problem.
>
> Well, this would be true if the transition range were always measured from
> the center of the squares at sea level - but most modelling tools place the
> transition range measuring point at the average of all of the vertices.
> Hence, if square B had a *REALLY* high mountain in it somewhere (and yet still
> be lower than A at some point on the boundary) then it's transition
> range would be measured to a point much higher in elevation than square A
> (which could be flat).
Yes. If I flied through such nasty terrain like Grand Canyon or
Bryce Canyon, I would find lots of big holes even with the trick.
But for normal mountain areas, it works quite well.
> So, either generate fill-in triangles for both up and down transitions, or
> be careful to force transition ranges to be measured relative to a point
> at sea level.
Filling in up transition is not a good idea because the vertical
triangles stand along the tile boundary like walls even when
neighboring tiles have the same LOD. On the other hand, down
transition patches are hidden underground in such cases.
To place the LODCenter at sea level reduces the effect of LOD.
> Another point to remember is that you want to force the surface normals of
> your in-fill triangles to point more-or-less upwards or else the sun-shading
> of your vertical in-fill triangles will produce nasty black blobs in
> the terrain when the sun is high in the sky.
Right. By forcing the normals to be the same as those of the original
tiles (I used per-vetex basis normals), the surface brightness had no
problem except that the normal continuity is lost at the boundary of
tiles with different LOD. This problem occurs regardless of usage of
the patch triangles.
When texture is mapped, artificial vertical stripes appear on the
patch triangles. This sometimes looks ugly, but using LOD blending
softens this problem.
-- Masanori Kakimoto mailto:kaki++at++nsg.sgi.com East Asia Technology Network, Tokyo Office c/o Nihon Silicon Graphics K.K. TEL:+81-3-5488-1852 FAX:+81-3-5420-2397 Voicemail:5-7465
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:52:33 PDT