Page 2 of 2
Re: Signed Cloud to mesh distance
Posted: Fri Dec 04, 2015 5:23 pm
by daniel
The PoissonRecon plugin is powerful but you need really clean normals to get a good result.
I'll take a look at your file when I'll have more time. How do you generate your input meshes? Can't you fix the face orientation issue upstream?
Re: Signed Cloud to mesh distance
Posted: Fri Dec 04, 2015 5:53 pm
by olivier.martin
I made all my analyses in R (except what I am doing with your software ;) ). So, I generate my mesh with an alpha shape 3d function of R ("ashape3d") and I export it with another function of R ("writeOBJ"). I haven't many control on what I can export with this function... and the format generate by the alpha shape 3d function has many thinks (vertices of triangles, information on edges...)... but anythings about the normals.
I'm' trying to fix the faces orientation issue with the function I use in R but I didn't find how to yet. I keep trying anyways.
For now a solution I found is to keep the method your propose to me but with less accurate meshes (with a high value of the alpha parameter for the alpha shape 3d function). Even if the meshes doesn't correspond exactly to what I want the re-orientation of faces and the calculation of PoissonRecon are quite good in that case.
Thanks for your time.
Olivier
Re: Signed Cloud to mesh distance
Posted: Fri Dec 04, 2015 6:07 pm
by olivier.martin
I think faces are not logically oriented (its just random) in the mesh I generate. You told me on your first answer that the orientation depend on the order of the vertices ("direct mode"). What is exactly this order ? maybe I can reorganized the order of the points of my triangles... !?
Re: Signed Cloud to mesh distance
Posted: Sat Dec 05, 2015 8:52 am
by daniel
If you look at the face from the 'front' side (from the outside) then the triangle vertices should be in the counter clockwise order (see
https://www.opengl.org/wiki/Face_Culling - Winding order).