Meaning Regitration RMS

[yvdnbroe]

Dear all,

What does the ICP Registration RMS actually mean? Could I assume it's a value for the systematic error of my dataset when I register two datasets of the same object?

What's the relation between the ICP Registration RMS and the C2C Mean (Quadric) distance/standarddeviation?

Thanks in advance.

Regards,

Yannick

[daniel]

Indeed the RMS is the square root of the mean square errors. Therefore it's pretty much equivalent to the standard deviation of the relative distances between the two cloud. The main difference is that it's computed on less points (I believe the number of points is displayed in the report).

It incorporates (at least): the noise of your data/sensor, the registration errors and the density variations of both clouds.

[yvdnbroe]

I calculated the registration RMS with all my points. I assume this a Nearest Neighbour calculation.
When I don't define a local model, i.e. Nearest Neigbour calculation, with the C2C distance, this results in a lower value than registration RMS.
The C2C report does not show the number of points. So which is more accurate?

[daniel]

If the ICP method doesn't show the number of points then it means it has used the default 'Random sampling limit' number.

In this case the distances RMS which does use all the points can be pretty much different. And considering which one is more accurate depends on what you expect ;)