Terrain Modeling Replies

New Message Reply Date view Thread view Subject view Author view

Andrew C Jank (acjank++at++aia.af.mil)
Thu, 28 Mar 96 11:49:51 CST

Hi,
     Thanks for the replies in response to my original terrain modeling
problem. I elected to implement building "walls" to fill the cracks
downward for the lower resolution tiles, and it looks pretty good. Since
I am limited by the rendering speed of an Indigo XZ, I limited the size of
the walls by forming independent triangles to cover only those places
where a crack would show (building ~75% fewer triangles than a full wall
mesh would require). The problem with this method is that two tiles that
differ in resolution level by more than one will show additional cracks
(that the pre-computed, simplistic walls were not designed to
accommodate).

     I also checked out the documentation on TerraVision (thanks for the
reference) but I did not think that method of "averaging out" the
elevation of the skipped points in the higher resolution tiles would work
for me. In my app, I know the extents of the terrain, so I load in the
appropriate tiles and build the LODs/geodes/geosets ahead of time, not on
the fly like TerraVision seemed to do. For my tiles, the averaged
elevation would work fine for adjacent tiles of equal or lower
resolution, but the same tile would be presenting averaged elevation
values to an adjacent higher resolution tile. The higher resolution tile
would be expecting a full complement of accurate elevation values (so it
could average its skipped points and interface properly), BUT it is
actually interfacing a set of values where part of the values have been
averaged, ruining the neat tile interface, developing cracks, etc ...
(I understand how their method could work when the vertex lists, geodes,
etc are built on the fly, but it really does not seem to solve a problem
where the geodes are built ahead of time).

     Anyhow, what I have should meet my needs until we get funding for
Performer 2.1 and can use the new terrain building features.

Andy

New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:52:36 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.