i am working recently on a script to make quadrics surfaces on blender, i got to go from a point cloud to a mesh without face using a script (using scipy and a voronoi graph).
now i am willing to fill the tiangles that have their edges already built. If i use a simple F from the interface selecting the whole mesh, it doesn’t do very well…
My first question, how do i get the exact same result, if i copy paste the command in the script, i don’t get the same behavior, eventually someone encountered that before
My second is, is there an already existing tool that would fill my triangles without holes or those weird overlapping faces.
My third, does someone have a way to either detect or automatically clean these overlapping faces? (the merge by distance/remove doubles was done already)
I was wondering why such a bunch of triangles and why do you have to go from point cloud to mesh and don’t build up the surface?? Maybe for this task something like MeshLab is more suited? And if we speak of:
then why not use something like the Extra Objects addon using Add → Mesh → Math Function… ?
The terms Dxy Eyz Fzx as in the equation of the section Euclidean space can twist the surface quite far what it would look like without it, they are not really generic.
I don’t know any X(u,v) Y(u,v) Z(u,v) equation that can match any quatric equation, which is necessary if we want to do build those easily. So the Extra objects add-on, which uses that can’t create such surfaces without it (i like the add-on though, i didn’t know about it)
It is indeed exactly what meshlab does that i want to reproduce, but i will remain in blender as much as possible as i don’t have one quadric i want to build but a tool that can make one.
The mesh could and edges you see above are made with a script, so far i randomly place 1k or 3k vertices on the surface and it makes the edges using a 3D voronoi graph, it’s the furthest i have gone, filling faces between the edges that has smallest angle at each vertices looked promising but isn’t as good as i hoped.
I didn’t know the convex hull function, i kind of coded an equivalent which takes a ICO sphere bigger than the surface and shrinkwrap it at closest vertex, but it has issues with concave areas.
I don’t really know where to look for addons, but i didn’t find anything so far.
Now i only want to make a surface that will be remeshed later on to project it back on the theoretical surface, so i can still afford a bit of roughness
So far the generation is random, if i can get to fill those faces, i will change the vertex density where needed.
i didn’t choose columns assuming we don’t know how it will look like in advance and there are areas where i don’t expect it to work easily as i don’t know in which direction make the “grid”. maybe i should step back in this direction.
i use the one given at the beginning of the Euclidean space section of the quadric wiki page given by okidoki. You can just set two of the xyz, and solve the third one
I also used a bmesh, don’t know what is the old thing.
Yes i use the equation to get the third coordinate (set X,Y get Z, set X Z get Y…), complex values are to be discarded, only keep the reals ones, if you only have complex numbers it means that your quadric doesn’t cover the range you proposed. that depends on the abcdefghij you gave, for example if you gave it a radius 1m sphere, the range you gave won’t give anything
I didn’t stop it yet, just not entirely focusing on this part of the code, waiting for new ideas
