CloudCompare forum

Hi,

I am in a debate with my boss around the functionality of the 3 (or 4 now?)-point alignment tool. We use this tool to georeference our scans by aligning them to a set of control points that we have xyz coordinates for. My understanding of the tool is that at most it will scale up or down an entire point cloud to best match the reference points selected. My boss is arguing that the software is using a TIN created from the points selected (to be referenced) that warps the point cloud in some way based on these triangles to fit it to the reference points.

Is this correct? I feel this couldn't be true because despite the fact that a scan is created in arbitrary space, it is self-referenced no? Like if I'm attempting to tie down a perfect pentagon to four even points and one point with a drastically different elevation, the shape of the pentagon point cloud will not be altered to better fit the points, it will simply be translated/rotated to best fit the five reference points based on averages (while maintaining the integrity of the initial cloud)? Is this correct? Sorry this is difficult to explain...

Thanks in advance!

I believe you are correct. By default, the 'Align' tool will apply a single rigid 3D transformation (= rotation + translation) to the 'to-be-aligned' cloud so that the 3, 4 or more points you've selected on this cloud match the equivalent points in the 'reference' cloud as best as possible. There's a mathematical optimization behind that will guarantee that the sum of the square errors between all the pairs is minimal. If you only have 3 points, you could see that as matching 2 triangles as best as possible (but without altering their shapes). With more points, it's harder to use the triangle analogy anymore, as there's no explicit triangulation involved, and no local deformation/stretching.

You have the option to allow the tool to change the global scale of the 'to-be-aligned' cloud as well (the 'adjust scale' checkbox). In this case, the tool will also be able to scale the whole cloud in addition to translating and rotating it. But once again, this is a global scaling, not local scaling.