Finding the central 'skeleton' curve in a polysurface

Hi all, I'm having trouble solving something. I'd like to find the central 'spine' in a polysurface (as in the image). I've generated this example from a hex grid where that skeleton is already known, but is there any way to extrapolate such a curve when one only has the polysurface, as in the attached GH file? Any help/advice would be appreciated, thanks!

Attachments:

Hex grid - help.gh, 23 KB

Replies to This Discussion

Permalink Reply by Laurent DELRIEU on March 15, 2016 at 1:55am

Medial axis ?

Permalink Reply by peter fotiadis on March 15, 2016 at 4:02am

Well ... this is not the easiest thingy imaginable (general case, not that one).

So for this specific case get a 5 minutes job that undoubtedly qualifies as the most stupid code even known to mankind (anti-Nobel on sight, for sure).

PS: offset used is the wrong thing to do (convex stuff outwards > requires an other approach, but anyway)

Attachments:

Hexgridhelp_ultra_stupid_V1.gh, 136 KB

Permalink Reply by Alberto Senigagliesi on March 15, 2016 at 8:58am

Hi Andrew

is this what you want to reach

Attachments:

Hexagrid-.gh, 5 KB

Permalink Reply by Nik Willmore on March 15, 2016 at 9:50am

This project contains a good 2D surface-with-holes medial axis calculator:

http://www.grasshopper3d.com/photo/medial-axis-determination?contex...

However, it's way too slow except on a subset of your grid if I edit the middle out and make but a portion of it.

It also suffers from non-straight outer lines since the Voronoi strategy fails to create many cells along the outside.

Attachments:

Medial Axis TEMP A.gh, 23 KB

Permalink Reply by Nik Willmore on March 15, 2016 at 10:20am

Ah, by merely adding more points along the long outer boundary curve, I avoid the curvy artifacts:

This lets me use many fewer points along each curve, too.

Can I now do your whole array thus? The slow step is searching for the right lines to erase, so certainly I can get the lines, but can I isolate them? Smoothing them out is another matter. Your filleted corners also results in little extra segments of a true medial axis.

And it only takes a few seconds.