Georeferencing

[armstrong_geology]

I have seen the great videos on georeferncing in Cloud Compare that I am interested in using. I had some questions however;

I have a point cloud of a rock outcrop derived from Photosynth (photo synth.net) that I would like to compare with a high accuracy GPS survey point cloud of the same outcrop. I am testing if photo synth could replace the GPS survey, should it be accurate enough. To test the accuracy, I would like to georeference the photo synth point cloud to the accurate, but sparse GPS survey cloud. My question is, if I reference the photosynth cloud to the GPS survey cloud as instructed in the tutorial videos, will the photo synth point cloud's xyz values be updated to their real life coordinates when exported from cloud compare? I need these coordinates to be as true to real life as possible so that I can calculate the error associated with this methodology.

I couldn't be more excited to have discovered cloud compare as ArcGIS is not performing well with this task. I am compelled to exploit your expertise! Any advice would be greatly appreciated!

[daniel]

Hi,

I'm not sure about which video tutorial you are referring, but I guess it involves using the point-pair based alignment tool? If yes, then the coordinates will actually be modified and you'll be able to export them as you expect (i.e. in the GPS world).

However, if the GPS coordinates are very large then CC and OpenGL won't like it at all (see this post for instance for more information: viewtopic.php?t=68).

Therefore you should use the following trick:
- shift all your GPS coordinates by the same big constant values (i.e. the largest digits that are common to all of them and for each dimension--> e.g. if the X coordinate is around 567 987.9989, you can typically and easily remove 567 000 to all the points). The idea is to work in a 'world' with rather small coordinates (lower than 1000 would be the best but 10 000 or 100 000 is acceptable).
- afterwards you'll have to set the shift you've just applied to the GPS points as 'Global shift' ("Edit > Edit global shift") on the registered cloud. Warning: this time you'll have to apply the 'inverse' shift (i.e. '-567 000' instead of '567 000'). When you'll save the cloud, CC will apply the (inverse) shift so that the written points are expressed in the global world. This mechanism is in fact used internally by CC when importing clouds with big coordinates (it applies a shift on all the points at loading time, and it applies the inverse shift when saving the cloud). This way there's no loss of accuracy and OpenGL can display the clouds properly.

Sorry for all the manipulations. Updating the Point-pair based align tool to do this automatically is in the TODO list... but it's not ready yet ;).

[txrpls]

Hello all,

I am completely new to CC and have not downloaded the program yet. I was told by a friend that CC could be used for filtering ground data from aerial data (las). While doing a lot of reading on CC I came across this thread. If I am understanding correctly, CC will not be able to use State Plane coordinates Texas South Central US Feet which are x=2,000,000 range and Y = 13,000,000 range. Is my assumption correct?

Thanks,
Mark

[daniel]

In fact if it's only for cleaning there shouldn't be any problem: CC will just shift the cloud when loading it so you won't see its real coordinates while manipulating it. CC will eventually apply the inverse shift when saving the cloud so it should be 'transparent' for you.

This shift thing is just not yet implemented everywhere (and the point-pair based alignment is not yet ready). But cleaning should be ok.

[muzo583]

Dear all,

I'm wondering sth about georeferencing. Is there a possible to create and detect circle from target shape in point cloud? And if it is possible with ransac algorthim, can we use the center of this primitives for georeferencing?
I was using it in Polyworks, choosing target(or drawing polygon which includes shape) in point cloud and than it finds the center of chosed points (possibly aritmathic means of x,y,z) and create a single point for transformation (also georeferencing). I think it will be more accurate for transformations.
Any idea?

Thanks.

Mustafa.

[daniel]

It is planned to add semi automatic sphere detection for the Point-pair based alignment tool (which also let you define Ground Control Points manually, and thus georeference a cloud).

[daniel]

A little follow-up on this subject: the latest beta release (2.6.1) has a new feature for 'Point-pair based alignment': it can detect sphere where the user clicks.

See the updated documentation here:
http://www.cloudcompare.org/doc/wiki/in ... itle=Align

And additionally there's a new Tools > Fit > Sphere' method.

[Ptaszynka]

I couldn't be more excited to have discovered cloud compare as ArcGIS is not performing well with this task. I am compelled to exploit your expertise! Any advice would be greatly appreciated!