Print preview, LRUDs on vertical caves

[kevin dixon]

3D Modelling packages to consider for ideas :

MeshLab
CloudCompare
PointCloud LIbrary http://pointclouds.org/
Scanalyze http://www.graphics.stanford.edu/software/scanalyze/
3DTK (3D Toolkit) http://slam6d.sourceforge.net/
GNU Triangulated Surface Library http://gts.sourceforge.net/index.html
Robert Schneiders List http://www.robertschneiders.de/meshgeneration//software.html

Point Cloud Library may be a good start point.

With mine software such as Vulcan, you manually create a set of parallel cross sections which are then used to create the mesh. Not ideal because of the point interpolation and decimation of dense data. 

Meshlab will do complex shapes but takes a lot of effort trying different decimation scenarios, meshing algorithms and parameters. 

Cave Survey is a very sparse point cloud problem which suggests a solution specific to cave survey rather than using general point cloud algorithms.
Perhaps use the defined cave centrelines and blow up a tube ballon to fit the points/splays/LRUD to define the tube surface triangular mesh. Do this for each set of centrelines and then do a secondary operation to join all the meshes with a check that the junctions do not intersect splay/point/LRUD rays. 
Large chambers could be a problem if a closed traverse has been done round the edge as this might result in a torus shape. Might need some parameter to be set for the tube mesh joining to create a large chamber from adjacent tubes.

Hope the above is helpful.

Kevin Dixon

________________________________
 From: Mateusz Golicz <mateusz.golicz at pza.org.pl>
To: survex at survex.com 
Sent: Monday, 15 July 2013, 17:52
Subject: Re: Print preview, LRUDs on vertical caves
 

On 07/03/2013 02:59 PM, Erin Lynch wrote:

> We have a number of large shafts where we've recorded passage dimensions
> along four different bearings (i.e. 045,225,315,135) instead of using standard LRUDs.
> It would be helpful if survex had an option to enter this sort of data

I would say this is highly related to constructing not tubes, but rather "bubbles" from a bunch of splay shots in arbitrary directions, as resulting from surveying using the paperless kit (distoX + PDA).

I can commit to code it (well, maybe not in a few days)... but I am at a loss in search for a reasonable algorithm to do the job.

I would prefer not to resort to heuristics like interpolating a cross-section at every station and joining these cross-sections into tubes (what about junctions, halls etc...). Instead I seek something universal, that would work reasonably well for LRUDs, readings such as yours (passagestation045225315135), as well as for a few dozen of splays from the paperless kit.

Simple convex hulls are not good enough, they behave oddly for plain LRUDs (ceiling, floor, walls wooble). I have considered building an union of convex hulls for each centerline shot (instead of each station), but that's still not perfect.

I have also played a bit with MeshLab, but I noticed that when computational geometry scientists say something on reconstructing volumes from "sparse sets", they mean point clouds of 100s or 1000s. And we have just a few per every station (4 - LRUD or maybe 10 - 20 in case of distoX).

I am writing all this in hope someone here can point me to some resources/ideas on this subject?

Except 3d modelling, what appeals to me is exporting such bubbles (tubes) into SVG together with contour data - in order to get halfway to a neat overview map of the area, similar to this: http://pliki.jaszczur.org/kongres-xy.pdf

Cheers!
Mateusz

-- Survex http://lists.survex.com/mailman/listinfo/survex