Basic theory

[wharfbanger]

Hello, I am new to CloudCompare and have a basic question. I am using CloudCompare for change detection (vegetation, subsidence) using a time series of aerial survey photogrammetry point clouds.

My understanding is that the fine registration function aligns two clouds in the X and Y. Z is not aligned because the cloud/cloud distance function outputs the Z difference. Is this correct?

Also, my point clouds were georeferenced using an on-board GPS . This means there may be some distortion in the point clouds due to the inaccuracy of the GPS, particularly the Z coordinate. Any advice here? E.g. rebuild point clouds without GPS coords? Perhaps I can focus on a small sub-area where distortion is likely to be minimized? I have also considered rebuilding the point clouds after first editing those GPS coordinates where the Z coords is obviously inaccurate..

I appreciate my issue may be more related to the accuracy of the initial photogrammetry processing (rather than CloudCompare), but would greatly appreciate advice here if possible.

Thanks,
Mark

[daniel]

Nope, the 'Fine registration' algorithm is by default working in '3D' (all dimensions are used without any difference - the distance used is the 3D euclidean distance). You have an option to limit the displacement along one or several dimensions however (but it shouldn't be enabled by default).

For the second question, it doesn't think really simple ;). I'm not an expert for this, but I'm sure other users have some experience with this.

[wharfbanger]

Hi Daniel,

Thank you so much for your reply. However I am a little confused by an apparent paradox. For example, I have two point clouds of a town (before and after subsidence). Subsidence is suspected to be forming a circular topographic basin in the central town (say 75% of the point cloud surface area). How can we detect subsidence when the fine registration step is "removing" the subsidence?

Thank you,
Mark

[daniel]

In this case, you shouldn't use the moving parts for registration.

Clearly in this case it's better to use the Align tool (with as much pairs as possible). Otherwise, you can segment out the 'fixed' part of the aligned cloud and use only this one with the ICP algorithm. Then apply the resulting transformation matrix (copy it from the console) and apply it to the original cloud (with the 'Edit > Apply transformation' tool).

[wharfbanger]

Thank you Daniel.

That makes sense to me.

Best,
Mark