surfaces via point cloud

G'day there,
I'm trying to create a mesh of a seat (some curved surfaces, some flat), and
I was wondering if anyone had any tips on the best method. My end aim is to
import the model into HyperMesh for use in some dynamic analyses.
So far I've:
1. Used a FARO arm to scan the seat. This creates a very fine mesh of
triangles - far too complex for my purposes, so...
2. Used ModelMax (supplied FARO arm software) to perform a number of mesh
reductions (fill holes, weld adjacent triangles, resample points at greater
intervals). This seems much more an art than a science, and I'm still not
entirely happy with the result (ie, a much coarser mesh of triangles that
still look somewhat random).
The point is, I can almost *see* how I would like my seat mesh to be
arranged (ie, where I would put the 'simple' flat panels and the more
complex curves), but I don't know if it's possible to go from the complex,
random triangle mesh to this simple mesh.
Does anyone know of a standard 'technique' and/or tool that people use to
generate a clean mesh from a digitizer like the FARO arm?
Any help would be appreciated.
Cheers, Sven.
Reply to
Sven
Loading thread data ...
Look into GeoMagic Raindrop software. This does a great job of turning point clouds into surfaces. A well-defined surface can then be used to generate your mesh.
Reply to
TheTick
In all likelyhood the original part was designed using geometry, such as lines & tangent arcs, perhaps some more complex types. Try to think like the original designer. Just create what you suspect were his original controlling curves (based on your scanned data) and go from there. You will learn and may need several tries. Remember that your input data now has a tolerance made up of the original design & manufacturing tolerances PLUS the size of the 3D error from your "scan". You cannot know beyond that but can use it to modify/smooth your data. Anything inside that resultant 3D tolerance band is probably equally good ...
Just hope the guy was not a "stylist". Then all bets (as well as good design & engineering) are off.
HTH
Reply to
Cliff Huprich
Thank you, Mr Tick, this is the kind of thing I'm looking for. Looks (from their website) like a very intelligent suite of products.
Ta, Sven.
Reply to
Sven
Thanks Cliff, I'm actually working on this kind of method now anyway... I'm selecting points from my point cloud to manually define surface 'patches'. The seat is actually a baby capsule, so its geometry is slightly more complex than I may have indicated in the last post, but that just means this method will take a little longer... and besides, I'm learning so it can't be detrimental.
Cheers, Sven.
Reply to
Sven
Sven, You may have missed my point. Patches & surfaces are *usually* the result of operations, such as sweeping, of other curves, such as lines, arcs & splines, along other curves. Sweeping is just one example but the idea is that there are basic generator curves for most surfaces.
Constructing these, as "fits" to your point data (example: it looks like 500 points fall along an arc [and that that arc may control/generate a surface in some way] Construct an arc with three of them that covers the entire range). Trim the resultant arc to the other generating geometry for the surface(s) then use your geometry to generate the surface(s).
Compare the surface to the original point cloud, which should now (if all went well and you were a bit lucky) be within the tolerance I mentioned before.
I hope that this was a bit less muddled .
Using the point cloud directly it's self is bound to produce all sorts of nasty bumps & ripples in the surface as each point is only (somewhat randomly) within the tolerance band I mentioned.
HTH
Reply to
Cliff Huprich
Ah, don't worry Cliff, I understood your method, and when I say that I'm selecting points from my cloud to define 'patches', I mean that I'm selecting *a small number* of points, but these points define the large flat surfaces (SW terminology) on the baby capsule. Therefore I'm not introducing small bumps/ripples etc into my final surfaces, as I'm not literally using the whole cloud.
I guess the main difference is that I've chosen to make just these flat surfaces manually, and then 'join' my adjacent surfaces with fillets, rather than define the full curve of parts of the capsule and sweeping along these curves. I think that the geometry I'm trying to make validates my choice of using the flat surfaces > fillet method, but I still see the merit of your suggestion, and I'll probably have a go at your method as a comparison as well.
Cheers, Sven.
Reply to
Sven
Sven, Quite possibly. Nobody else knows what the part looks like .
General idea: think like the original designer. Once you do a bit of that you can usually do a fair job of knowing what you need and why.
I had to assume the worst ... LOL .... and I've seen worms ...
Reply to
Cliff Huprich
...Have a look at what Delcams CopyCad has to offer, it takes point clouds back to surfaces / geometry, with some handy tools.
Best,
JAG
Reply to
JAG
Also have a look at Raindrop's Geomagic, this seeme to be the be all and end all for handling point clouds.
Reply to
D. Short
There is no escape from the tolerance issue .
EXCEPT: The original designer usually likes *nice* numbers. Knowing if it was inch or metric helps.
Then you may have better luck.
IE, If inch and the point is .376 with a .002 tolerance you can probably bet it's really .375. This can help you quite a bit but I don't think that it helps programs that *must* fit to the exact input data points much. At best, they may have "weights". Feel free to correct me.
Reply to
Cliff Huprich

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.