Subject:
|
Re: Advice on hair pieces
|
Newsgroups:
|
lugnet.cad.dat.parts
|
Date:
|
Fri, 17 Sep 2010 04:55:12 GMT
|
Viewed:
|
24599 times
|
| |
|
|
In lugnet.cad.dat.parts, Philippe Hurbain wrote:
> Actually original parts only have triangles and have been merged by stl2dat (the
> 4th batch used a different concerter and no triangle to quad merge is done).
I noticed the 4th-batch part I worked on had a significantly better format. That
must be why. It was missing some edge lines, but Edger took care of that
handily.
> > To my knowledge getting the original file and re-converting isn't feasible, so
> > I'm looking into what it will take to write a utility that tries to adjust the
> > vertices to restore coplanarity.
> Interesting idea. A radical method is to split offending quads into triangles,
> this can be done automatically with Planarcheck or Datheader.
Heh, I would've thought my idea of a utility was more radical. I'm not entirely
sure how solvable a problem vertex adjustment is. The concept I'm working with
is this:
1) Set constraints on each vertex for how far they can go, which will be based
on the number of decimal places they have in each dimension (e.g., 2 decimal
places means ±0.005 of wiggle room)
2) Identify non-coplanar quads (with 1° tolerance).
3) For each such quad, take the average face normal of all possible triangles it
could be split into and use that for the "ideal" normal.
4) Take the average point of all the vertices in the quad to find the ideal
plane, and give each vertex a "tension" toward that plane. It should not try to
cover the full distance to fit the plane exactly.
5) (Theoretical) Divide each tension vector by the number of quads in which that
vertex is being adjusted.
6) Adjust the vertex positions, clip them to their bounding region, round them
off to 4 decimal points, and return to step 2 unless too many iterations have
elapsed.
> You could only grow in horizontal direction. Going from 24.5 to 25ldu means only
> a 2% size increase, I don't think it could be noticed.
It would actually be a 4% increase since the interior is 25 LDU across and it'd
be scaling up to 26. However I think the exterior of the part looks correct in
all details; it's just the interior that's the problem.
> I have sometimes carved a bigger hole (using Intersector,
> http://www.philohome.com/isecalc/intersector.htm), but that's a big job.
I think I might start by simply adjusting the interior vertices outward where
they're too far in, and making sure the resulting planes still don't intersect
the head.
> I have done a primitive for that:
> http://www.ldraw.org/cgi-bin/ptdetail.cgi?f=p/stud4p.dat
Ooh, that's perfect. I didn't see that in the primitives ref but it's probably
because it's still not in an official parts release. Thanks.
|
|
Message has 1 Reply:
 | | Re: Advice on hair pieces
|
| (...) On the opposite! splitting stupidly into triangles is radical! The method you suggest is much more sophisticated and interesting ;o) (...) Yep. have a look at discussion here: (URL) it would be nice to have a tool to adjust vertices (...) (15 years ago, 17-Sep-10, to lugnet.cad.dat.parts)
|
Message is in Reply To:
| | Re: Advice on hair pieces
|
| (...) Actually original parts only have triangles and have been merged by stl2dat (the 4th batch used a different concerter and no triangle to quad merge is done). That said, I had a look at m62696, there are only 5 non-coplanar quads, all within (...) (15 years ago, 16-Sep-10, to lugnet.cad.dat.parts)
|
12 Messages in This Thread:
  
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
