.NET integratable robust, fast solution for surface reconstruction from 3D point cloud

Question

We have a .NET 4.0 C# WPF, machining simulation application. Right now we are stuck since we have to generate a 3D surface (mesh generation) from a point cloud. We have been searching, goggling, testing different libraries in order to find a solution that is fast, robust, correct and that can be integrated to our application but we had no luck. 3D imaging is not our line of business so we don't want to get into coding this part. A commercial solution with an affordable price can also be an answer for us.

Until now we have tried:

1. VCGLib through MeshLab UI but the algorithm they use (Ball Pivoting) didn't give a good result.

2. A MatLab code based on PowerCrust algorithm ( MyCrust Robust ) and it gave a good result but we don't think that if we recode this code in C#, it will gave us the performance that we want. We need a solution that can be easily integrated to our project.

3. While searching we have also came up with these sample applications . We don't know the detail of the algorithm used in these applications (details are in an academic paper) but it generated a result with so many missing meshes. But may be it can help to someone else.

Right now we are trying to learn CGAL and find out whether it can be an answer for us. But since installing Boost and compiling CGAL took two days for us we just started on working on it.

Can anybody offer us a solution that can be in help to us?

Thanks in advance

Answer 1

There is the G# library at http://www.ceometric.com It can do a surface reconstruction from triangular or quadrilateral wireframe data. This is not exactly what you have been looking for but it may be a first step towards a solution.

Answer 2

I am getting mu data from the kinect depth map and I found these papers:

• http://www-graphics.stanford.edu/papers/zipper/zipper.pdf
• http://www-graphics.stanford.edu/papers/volrange/volrange.pdf

They might be helpful if you have the Poisson algorithm in C# it would be great.