CGAL users discussion list

["Laurent Rineau (CGAL/GeometryFactory)" <>]

Le Wednesday, October 4, 2017 7:55:28 PM CEST Michael Bieri a écrit :

> Hi Laurent

>

> I investigated the problem by plotting the polylines. You can find

> increasingly more zoomed images here:

> https://www.dropbox.com/sh/lz8xmwhr3fg23qu/AAAZEpWnKzJyrjhNB006_-fVa?dl=0

>

> Points of the polylines are plotted in yellow, while the connecting lines

> are in blue. YES, there are so many points that the lines basically appear

> yellow. When zooming in, one can see that the polylines follow the pixels

> of the image. When zooming in even more, one can see that the pixel

> boundary is not just made up of one polyline, but of quite a number of

> points/lines. Why is this?

 

The 1D-curves of the domain are made line segments, and hyperbolic segments, and the polylines you plotted are an approximation of those curves:

 

- the approximation of a line segment is a line segment, with only two points,

 

- the approximation of an hyperbolic segment is a polyline with 10 or 11 points.

 

 

That explains what you saw while plotting the polylines.

 

 

> So...if the mesher actually wants to preserve ALL those features, no

> surprise that elements are getting extremely small. BUT I also see that in

> many parts of the mesh, elements along material boundaries are MUCH larger

> than the smallest 1D features. So I wonder - what's the rule, when is a

> polyline/feature exactly kept, and when does the mesher just approximate

> it?

 

The 1D-curves, represented by there polylines approximation, are then *remeshed* again, according to the geometric features of the demain, and to the size criterion given by the user. That is why the dentisy of points in the appromation of the polylines given to the domain has nothing to do with the actual density of the representation of those curves in the output mesh.

 

 

Would you mind putting on Dropbox the actual image you used, as well as the code? Or at least the file representing the polylines you plotted? I would like to verify if there is a small feature size in the curves (indenpendant of sampling of the curve).

 

--

Laurent Rineau, PhD

R&D Engineer at GeometryFactory http://www.geometryfactory.com/

Release Manager of the CGAL Project http://www.cgal.org/