Steve Chapman,
I can understand Steve Fareninos' skepticism of your ability to create easily
usable surface files from an STL or 3D point cloud. All software packages that
I have reviewed to date (not that I have reviewed them all) that claim to be
able to do this, only create surfaces of the triangular planes, therefore
creating huge databases that are essentially useless due to there huge size.
If you are making nurb surface models from STL or 3D point cloud data that are
manageable (either for modeling or machining purposes), I would be very
interested is seeing some sample surface models if they are available, and
knowing what software you are you using?
As a side note, creating a nurb surface model from 3D scan data (data from a 3D
scanner) is a different animal than creating one from an STL file. 3D scan
data is more or less a uniformly paced point cloud, while an STL file is a
triangulated/polygonized point cloud that is optimized for the specific
object. Therefore, the point in an STL file are not uniformly spaced. For
example, a surface on the object that is a plane, or close to being one, will
be described with less than 100 points in an STL file of the object, while it
will be described by several thousand points in a 3D scan file.
Therefore, any attempt to create a nurb surface of an object from an STL file
will miserably replicate any flat or nearly flat sections of the object since
there are so few points to describe the object in this area. Consequently,
unless I am missing something, no matter how robust a surface to point cloud
fitting algorithm you use, much work will be needed to correctly capture any
flat areas on the object.
I am very interested in better understanding the capabilities of you methods.
--B. Matthew Michaelis - Product Development Engineer Compression, Inc. 25242 Arctic Ocean Drive Lake Forest, CA 92630 (949)586-5875 http://www.carpediem.com Matt.Michaelis@carpediem.com
SCat3D@aol.com wrote:
> <<But my point was that, with typical STL file of say 50,000 facets, you get > an enormous database that's extremely difficult to work with. >> > > Try 5,000,000 points of data from a Cyberware scanner tuned into 1-50 or so > nurbs and solidified in Pro/E. Easy to work with. Check out my company in > the Sept. User Focus in CGW. See the doggy? 12 surfaces. > > For more information about the rp-ml, see http://ltk.hut.fi/rp-ml/
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 - 22:46:56 EEST