Brep from outer points in a "pointcloud"

I would like to create one (bubble) Brep from the outer points in a "pointcloud"/group. Does anyone have an idea how it could be done? :-)

Attachments:

points.gh, 8 KB

Replies to This Discussion

Permalink Reply by Michael Pryor on March 10, 2013 at 10:30am

getting the dip down will be difficult unless you use some marching cube meshing maybe with millipied. Otherwise try Starlings 3d convex hull.

Permalink Reply by alex jenter on March 10, 2013 at 11:04am

Im agree with Michael, what you want is referred to as (nurbs)surface reconstruction, and can sometimes be very tricky, even in specialized Softwarepackages that ship with 3d scanners. Solving it via remeshing (marching cubes) would be a lot easier and a more general approach. SL 3d convex hull might work in some cases but you would loose all concave or hole details.

For finding the outer points, you can sort them by z coordinates into groups, perform delaunymeshing (group by group) and run kangaroos naked vertices on the resulting meshes.

Permalink Reply by Ola Jaensson on March 11, 2013 at 6:52am

Or sort them by x coordinates and work with the yz planes to avoid 1 to 2 curve lofting.

Or maybe that wouldn't give you a useful brep?

Permalink Reply by alex jenter on March 11, 2013 at 7:33am

True for this point cloud, but as i understand Thomas is looking for a more general for any point cloud. Then depending on topology you run into other problems, like what happens when then point cloud closes up at the top side and forms a torus, you would get the 1 to 2 curve lofting (Y branch) in x or z direction.

Permalink Reply by Thomas Perkov on March 10, 2013 at 11:09am

Thanks for the answers :-)

I'm thinking of sorting into branches by z values, point group and using the convex hull component, fillet and then loft.. Is there a way to loft one curve to two curves?

Permalink Reply by alex jenter on March 10, 2013 at 12:00pm

For this one point cloud, you could try splitting the topmost "single curve layer" into two curves, then loft the "two curve per layer Stack" and the "one curve per Layer stack" separately. But that is not very elegant nor pretty nor a general solution to your problem. already tried marching cubes (from millepede)? Combined with Weaverbirds smoothing components you could achieve very pleasing results on a mesh based approach.

Permalink Reply by Andrew Heumann on March 10, 2013 at 1:36pm

Alex's suggestion is a good one. To set it up right you'll probably want the sample points to be on the same spacing module as your starting points. Here's an example:

Attachments:

points.gh, 22 KB

Permalink Reply by Michael Pryor on March 10, 2013 at 2:27pm

are you using r5, cant seem to get milli to work on r5

Permalink Reply by Thomas Perkov on March 10, 2013 at 2:39pm

I'm using r5 64 bit, and the def. works :-)

Permalink Reply by Michael Pryor on March 10, 2013 at 3:20pm

hmnn for some reason it wont show up in my r5 grasshopper, have downloaded and installed the r5 milli. Never had this issue before. It is working in rhino 4.

Permalink Reply by Marios Tsiliakos on March 10, 2013 at 3:32pm

I think this is an issue with the download link. a friend of mine couldn't make it either, so I send him the files from the very first download link of Millipede, and it was working. Rhino5 32 bit. maybe we should report it to the group if there is a consistency..

Permalink Reply by Michael Pryor on March 10, 2013 at 4:46pm

I think this may be it. The download instruction file is instructions for r4, even in the r5 download.