Delaunay Mesh VB.net grasshopper

Hi,

I need to make a Delaunay mesh within a VB component. I have done the following and the mesh seems pretty nice; except it acts strangely when I try to extract its normal vectors and then they are null (it may also cause Rhino crash!). Does anyone know what is wrong with this mesh?

Private Sub RunScript(ByVal x As List(Of Point3d), ByRef A As Object)

Dim N2L As New grasshopper.Kernel.Geometry.Node2List(x)
Dim DF = grasshopper.Kernel.Geometry.Delaunay.Solver.Solve_Faces(N2L, 1)
Dim Faces As New list(Of MeshFace)
For Each DFace As grasshopper.Kernel.Geometry.Delaunay.FaceEx In DF
Faces.Add(New MeshFace(DFace.A, Dface.B, DFace.C))
Next
Dim DM As new Mesh
DM.Vertices.AddVertices(x)
DM.Faces.AddFaces(Faces)
A = DM

End Sub

Ps. the reason I am introducing vertices in this way is that I want the mesh be a 3D mesh, just like the Delaunay mesh generated by the native component of GH.

Thanks,

Pirouz

Replies to This Discussion

Permalink Reply by David Rutten on December 3, 2013 at 6:03am

You should probably specifically tell the Mesh to compute vertex and face normals.

Mesh.Normals.ComputeNormals()

David Rutten

david@mcneel.com

Permalink Reply by Pirouz Nourian on December 3, 2013 at 6:39am

Yes! that's it! Thanks a lot!

But as Richard has also noticed, it sometimes puts out self-intersecting results. I wonder what is the role of that jitter value in there? I tried changing it and checking the differences with your Delaunay component, I thought it should be set to 1 but this was just a wet finger method. Could you please shed some light on this matter? Is that because a Delaunay triangulation is not unique and by changing the way points are sorted every time we get a different one?

Permalink Reply by David Rutten on December 3, 2013 at 7:20am

If there are more than 3 points on the same circle then the delaunay triangulation becomes non-unique. It shouldn't lead to problems, but just in case it does, the jitter allows you to randomly move points a bit before triangulating the point-set.

Use jitter=0.0 unless you run into problems.

David Rutten

david@mcneel.com

Permalink Reply by Pirouz Nourian on December 3, 2013 at 7:26am

Thanks for the explanation. Could you please also suggest a way around the problem of the 3D mesh? How to compute the mesh according to a Grasshopper.Kernel.Geometry.Node3List(x), so that the Delaunay mesh becomes a 3D mesh? (I noticed that none of the grasshopper.kernel.geometry.delaunay methods support points of Node3list type!?)

Permalink Reply by Richard Schaffranek on December 3, 2013 at 7:47am

Use you script and just set the jitter value to 0.0.

Permalink Reply by Pirouz Nourian on December 3, 2013 at 10:41am

Yep! : ) Thanks Richard!

Permalink Reply by Richard Schaffranek on December 3, 2013 at 6:20am

Hallo Pirouz,

copy and pasted your code. I am not entirely sure but the resulting Delaunay triangulation seams wrong. You can explore that by using Populate2D, at least in my case there are overlapping faces...

Permalink Reply by Pirouz Nourian on December 3, 2013 at 6:43am

Hi Richard, yes. Thanks for pointing to this. I have noticed that as well. Any ideas why?

Permalink Reply by Richard Schaffranek on December 3, 2013 at 6:46am

I would guess that the vertices are resorted in one of these two lines, but that's just a wild guess:

Dim N2L As New grasshopper.Kernel.Geometry.Node2List(x)
Dim DF = grasshopper.Kernel.Geometry.Delaunay.Solver.Solve_Faces(N2L, 1)

Permalink Reply by Pirouz Nourian on December 3, 2013 at 7:09am

The thing is, when you use this method, you always get a 2D triangulation, as you are using 2D GH_points as vertices. This is why I have used the faces only and made a mesh with 3D vertices. And this is perhaps why it acts strangely sometimes as the faces have been computed for a set of non-elevated vertices and then I use the original vertices to define the mesh geometry. I wonder what is the right way of doing it!? because I could not find the Delaunay methods for 3d points.

Permalink Reply by Richard Schaffranek on December 3, 2013 at 7:23am

No perfectly fine way to do it, but if the jitter makes me guess that the nodes are resorted (for what reason I can not tell since I never implemented a Delaunay algorithm. But I would guess that this might be to overcome rounding errors.->ask David for more details (solved while typing)).

You can see it even better when using your algorithmen on a grid. If you set it to something else than 0 nodes with the same y-value are resorted and therfore you can not simply replace them with the nodes they where generated with.

RSS