Registration of point cloud of same region (From different times) with different orientation

[daniel]

Ouch... If you don't have common features in each scan that you are sure that they haven't move (like big rocks, etc.) then it will be very hard to get a reliable result.

If you do have such structures, then you should first use the Point-pair based alignment tool (http://www.cloudcompare.org/doc/wiki/in ... itle=Align) to get at least a rough registration, and then optionally the fine registration (ICP) tool (http://www.cloudcompare.org/doc/wiki/in ... ?title=ICP) to improve this a little bit (but it depends on how much 'static' points are overlaid (use the 'Overlap' percentage to give a rough hint).

And indeed, you can try to mesh the reference cloud (if there are not too much hole or complicated structures). Or use the M3C2 plugin.

[daniel]

- the overlap percentage is the main parameter of the ICP algorithm (at least in the recent versions of CC). You really need to set it up correctly when performing the automatic registration (especially in your case). Otherwise ICP might even make the registration worse than it was after the point pair based alignment

- you can indeed create a plane (either with the 'File > Primitive Factory' tool, or with the new 'Edit > Plane > Create' tool in the 2.8 beta version - it will give you much more control). And then you can use this plane to compute the distances to your cloud with the 'Cloud 2 Mesh distances' computation tool.

Note that the '2.5D Volume Calculation' tool also allows you to compute the volume relatively to a horizontal plane (you simply need to input the height of the plane).

[daniel]

When you compute the distances, it also creates a 'scalar field' (one value per point = distances here). You can color the cloud with these distances, show the color scale, define your own color scale, pick labels on particular points to see the exact distance, display the distances histogram, etc. After all CloudCompare is designed to do exactly that :D.

You can also export the distances to an ASCII file (this will give you X, Y, Z and the distance for each point). You can then easily import such a file in Matlab, R, Excel, etc.

[daniel]

Simply select the cloud (with distances) and save it as an ASCII file (select the right filter). You can name this file 'anything.csv' to make things even easier with Excel. Last but not least, when saving as an ASCII file you'll see a dialog: choose the 'semicolon' separator. This way you'll get a true CSV file that excel will load correctly.

[daniel]

Not sure to understand: the plane is created exactly where you specify it (the center is expressed in the same coordinate system as the cloud). Therefore if all your clouds are registered together, or if all your patches are sharing the same coordinate system (at least for the Z) then you should be able to use the same plane. I guess :D