Page 1 of 1
Registration Using Signed Distances
Posted: Wed Feb 23, 2022 7:25 am
by Lentep
Hi,
How does the ICP algorithm work with and without "use C2M signed distances" option? Is it somehow changing the RMS calculated?
I am trying to decide if this option is good for my case. Just curious as well. Read the tooltip but it leaves me with some doubts and I cannot make sense of the code by myself. Hope someone can help, thank you
Re: Registration Using Signed Distances
Posted: Thu Feb 24, 2022 4:57 pm
by daniel
So the idea is to consider "negative" distances or not when we filter the farthest points at each iteration. In effect it has an impact when you check the 'Enabled farthest points removal' option or if you use a partial overlap of less than 100%. In this case, the points with negative distances will always be kept in preference to those with positive distances. This will have the effect of 'pushing' the points below the mesh surface (= with negative distance), closer to the surface at the next iteration... Hence preventing the cloud from 'sinking' below the surface. As much as possible of course.
Re: Registration Using Signed Distances
Posted: Fri Mar 04, 2022 6:25 am
by Lentep
daniel wrote: ↑Thu Feb 24, 2022 4:57 pm
So the idea is to consider "negative" distances or not when we filter the farthest points at each iteration. In effect it has an impact when you check the 'Enabled farthest points removal' option or if you use a partial overlap of less than 100%. In this case, the points with negative distances will always be kept in preference to those with positive distances. This will have the effect of 'pushing' the points below the mesh surface (= with negative distance), closer to the surface at the next iteration... Hence preventing the cloud from 'sinking' below the surface. As much as possible of course.
That make sense, thank you Daniel. Love this tool :)