Page 1 of 1
Fine registration - similarity or affine transformation?
Posted: Fri Apr 03, 2015 11:38 am
by anna
Please could you confirm that the fine registration tool (ICP) uses a similarity transformation when scale is selected? ie uniform or global scaling as opposed to scaling plus shearing which occurs with an affine transformation.
Noted in the documentation the point picking algorithm is affine which makes sense. Just double checking fine registration is not affine as using to align photogrammetry models to TLS and then checking relative distances in photogrammetry models.
Thanks
Re: Fine registration - similarity or affine transformation?
Posted: Fri Apr 03, 2015 12:43 pm
by daniel
Both registration algorithms (point-picking or ICP) output a similarity transformation. The scaling is always global.
The only way to apply a transformation that doesn't preserve the shape is through the "Edit > Multiply/Scale" (where you can set a per-dimension scale). And of course with the 'Edit > Apply transformation' tool as you can input your own 4x4 matrix.
Re: Fine registration - similarity or affine transformation?
Posted: Fri Apr 03, 2015 3:11 pm
by anna
Ah, ok thanks! I should have read this more carefully in the documentation and noted it specifically applies to failing to invert the transformation.
Page 104... Advanced parameters
This tool can also determine the best scale factor between the two point sets. To allow the optimization of the scale parameter, simply uncheck the fixed scale checkbox.
This can be very useful if the two clouds have different scales and you don't know the scale factor between them (otherwise you could use either the 'Edit > Multiply/Scale' tool).
Warning: if the scale is not fixed (i.e. different from 1) then the 4x4 transformation matrix issued on completion of this tool will be affine. Typically, the Edit > Apply transformation tool will fail to invert it...