I'm trying to figure out if there is a good way of doing surface classification for a geometry such as the shape below. Since it's a triangulated surface I have the coordinates of all the vertices and the connectivity that desctibes the faces. From this I can calcualte the surface normals, also visible in the figure below. I'd like to the be able to classify the geoemtry into three surfaces - top, edge and bottom, but while it looks like three distict groups of surface normals to the naked eye, I'm not sure how to do this numerically with something like numpy or scipy. Because of the bend on each end the direction of the surface normals at the tips could easily be mixed up with the direction of edge normals or the bottom of the other side. Any tips on a robust way of achieving this would be greatly appreciated. Continue reading...
