Registration via Latitude/Longitude/Elevation (or SPCS)

[lgeis]

Greetings from Boise!

I produce Point Clouds from my TLS (FARO X330) in SCENE; in this case, I'm "shooting" residential yards.

Specifically, I shot the yards once and then again several months later. Each time I had several 140mm spheres laying around and had their centerpoints surveyed (Lat/Long/Elev).

I use FARO's SCENE (7.x) software to assemble the Point Clouds and export .e57, et. al. I can obviously assign the spheres in the Point Cloud positions (again, Lat/Long/Elev). I'm uncertain as to whether or not that information survives the export to .e57, by the way.

I'd like to import a point cloud from two Point Clouds (shot several months apart) and have them register based upon the location of the spheres rather than the point constellations (via ICP). Of course, the positions of the spheres vary between the two shoots...

Can't find anything that might even touch on this in the manual or the online forum. Does anyone know how this could be accomplished in CC (my goal is to still apply the Cloud to Cloud Distance Tool to produce a visualization).

Many thanks in advance.

[daniel]

I don't know how Scene works exactly, but I would assume that once a point cloud is geo-localized, it should be exported in its new coordinate system (especially with E57).

If it doesn't work, then you still have the option to redo the geo-localization with the 'Align' tool? (http://www.cloudcompare.org/doc/wiki/in ... itle=Align).

[matknaak]

Hi Daniel & Igeis

Faro Scene certainly allows to export georeferenced point clouds no matter which format. If I understand Igeis corectly, he wants to detect the changes between the two point clouds by having them georeferrenced previously using surveyed spheres, and then calculate the difference in CC. The problem I see though is that most likely the changes he wants to detect are smaller or at the same order than the accuracy of the surveyed spheres. In which case the correct location of the point cloud to a referrence system is irrelevant. It would be much better to allign the two point clouds in CC using identical points from the two point clouds which have not changed. This can be achived using ICP, but I am unsure about the exact procedure and how many points are to be used for alignment.
cheers

mat

[daniel]

If you want to use ICP, then how to use it will highly depend on the overlap area, the density of the reference cloud, etc.

For a first try, the main parameter is definitely the 'overlap' one (you can estimate it roughly, it's only 10% increments).

[lgeis]

Thank you all very much; I'll put the reference Point Clouds to the test when that becomes a part of the current assignment and report back on successes...