Re: Constructing pfASD
Steve Gifford (sjg++at++terrex.com)
Wed, 15 Apr 1998 16:28:27 -0700
Dan Brockway wrote:
>
> Lee Willis wrote:
> >
> > ----------
> > From: Thom DeCarlo <trdecarlo++at++tasc.com>
> > >> Lee Willis wrote:
> > >> From: Dan Brockway <db++at++ParadigmSim.com>
> > >>
> > >>> If, on the other hand, you have a specific terrain surface at the
> > highest
> > >>> LOD that correlates with other databases, then MultiGen's CAT is of no
> > >>> value.
> > >>
> > >> No argument. But the problem goes deeper than that. If you have a
> > specific
> > >> terrain surface at the highest LOD that correlates with other databases,
> > >> ***ASD*** is of no value. (unless you're database happens to be a
> > regular
> > >> grid)
> > >>
> > >> You cannot construct an ASD from an arbitrary triangular mesh. It just
> > aint
> > >> doable.
> > >>
> > >
> > >I'm sure you will correct me if I am wrong, but since we are talking about
> > >starting from the raw source (DTED type data), we do have a regular grid.
> > The
> > >triangular mesh needn't be generated until much later in the process. It
> > seems
> > >to me that there should be a straight-forward, though time-consuming,
> > process
> > >that looks at each point in the highest resolution set, finds the nearest
> > >point in the next lower resolution set, and identifies that as the "morph
> > to"
> > >point.Since you say this is hard, there must be something I am missing.
> >
> > You and Dan are talking about different scenarios.
> >
> > Dan is talking about having a preconstructed polygonal database at the best
> > LOD,
> > and then trying to derive coarser LODs from that set of polygons -- can't
> > be done.
> >
> > You're talking about starting from elevation data -- that *can* be done.
> > Multigen CAT
> > does it, or you can write you're own.
> >
> > In very broad strokes, here's how:
> >
> > 1) construct a tesselation to be your base (coarsest) LOD. This can be
> > something as
> > simple as two triangles to form a rectangle, or you can use standard
> > terrain algorithms
> > (e.g. Delaunay) to construct an irregular mesh.
> >
> > 2) For each subsequen LOD:
> >
> > 2a) Using the previous LOD as a base, choose at most one new point to split
> > each
> > existing edge. You'll need to decide how to choose a new point. You can
> > use
> > something simple like 'worst vertical error within the neighborhood of the
> > edge'.
> > (i.e. compare the polygons to your elevation grid, pick the elevation
> > point that falls
> > furthest off the polygonal surface)
> >
> > More complicated algorithms are of course also possible. (e.g. ones which
> > may
> > weight some points more than others)
> >
> > 2b) Re-tesselate
> >
> > ----
> >
> > You'll notice that this algorithm doesn't start with all the points and
> > choose which ones to throw out.
> > It starts with a few points and then decides which new points to add. This
> > is what I mean by doing
> > it 'top-down' rather than 'bottom-up'. It's an insertion algorithm rather
> > than a decimation algorithm.
>
> So, if one modified this algorithm to insert only points from the source
> TIN instead of from the source DEM, one could arrive at the source TIN!
> This may not be straight forward, but possible.
You'd arrive at another TIN with the same set of points, but not
necessarily the same set of edges. For the application where the source
TIN must be matched exactly this still wouldn't do it.
-Steve Gifford
Terrex Inc.
=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
Submissions: info-performer++at++sgi.com
Admin. requests: info-performer-request++at++sgi.com
This archive was generated by hypermail 2.0b2
on Mon Aug 10 1998 - 17:57:15 PDT