Re: ATTN: Bert VandenBerg RE: stl file comparison

From: Bert VandenBerg (bert@hymarc.com)
Date: Thu Jun 18 1998 - 19:48:11 EEST


Monica & Glenn Whiteside wrote:
>
> Bert:
>
> The problem with polygonization of a scanned point cloud is that any
> outliers (stray points) can severely distort the STL model. Does Hymarc -
> Hyscan software filter out these outliers when scanning?

The design of the camera system is such that there are few outliers
:-).
Only on shiny surfaces do outliners appear. These surfaces are typically
treated with a airborne powder or other dulling spray.

> Does Hymarc - Hyscan have a similar control to
> use when creating STL's? Say a certain control distance from the point
> cloud?

The polygonization tests the distance of triangle-sides created, and
does not
keep triangles with especially long sides. This automatically ensures
that
"good" triangles are kept.

Note that a polygonized scan file still needs to be compared to the
model
for inspection results. A program like surfacer's verdict,
Innovmetric's
polyworks, or your internal tool is needed to do this registration,
and comparision.

> You said that the points defining the polygonization are
> associated/projected to design surfaces only if they are close enough.

My point was that a best fit that matches points only by distance is not
matching points as well as one that pays attention to both distance and
surface normal. If the fitting program checks that a point on a
(polygonized)
scan has approximately the same surface normal as the surface to which
it
is being matched, then there is a better chance that the points are
being matched properly. Consider the a surface with a hole. If the
measured points (letters a,b,...) are not polygonized, then points
d,e, will be matched to the wrong part of the left side of the hole
(if the pass the checking distance), while points e through k will be
matched to the wrong side of the hole. By polygonizing and checking
the surface normal as well as the distance, the fitting and the
inspection
result is improved. The maximum point to surface checking distance
can be increased, and the points still properly matched to their
nominal surface.

    a b c d e s t u v
                       f r
------------\ g /-------q---
             \ h /
             | i | p
             | j | o
             | k |
             | | n
             | l | m

-- 
Bert van den Berg                   E-mail: bert@hymarc.com
Hymarc Ltd.                           http://www.hymarc.com
35 Antares Drive                        Tel: (613) 727-1584
Ottawa, Ontario  K2E 8B1                Fax: (613) 727-0441

For more information about the rp-ml, see http://ltk.hut.fi/rp-ml/



This archive was generated by hypermail 2.1.2 : Tue Jun 05 2001 - 22:45:59 EEST