M3C2 and volume calculation

[Adam]

Hello,
I have just some questions about the M3C2
Is statement below correct?

The M3C2 algorithm can be broken into two steps: point normal estimation and difference computation. Users may specify if local point normals are calculated or if normals are fixed in either the horizontal or vertical direction. Horizontal point normals allow true horizontal erosion rates to be calculated from M3C2 analyses, while vertical normals allow M3C2 data to be used for strictly vertical erosion and aggradation measurements

and which one to use to detect changes vertical or horizontal?
What does core points mean?
Normal and projections keep changing when I use guess params, I don't know when to stop. is the normal values should be higher than the projection values ? for example: I get (diameter = 1.525847 for normal) and (diameter = 0.762924 for projection) if I click on guess params again I get diameter = 0.762924 for both and again (diameter = 0.381462 for normals) and (diameter = 0.762924 for projection)

Volume calculations
For the steps is 0.35 ok?
Cell height: I select average height?
Empty cells: I select leave empty?
Projection dir: I select Z?

and lastly, for the final overlap in ICP, Is it wrong to use 100%? I normally use 80% what happens if use less?
and what does enable furthest point removal do? should I use it?
I always ignore adjust scale?

Thank you

[daniel]

So many questions ;)

First I have one: have you read the wiki documentation and ideally the associated article?

Otherwise:
1) Whether to use the 3D normal, or the horizontal part, or only vertical ones depend on your assumptions on how the movement / changes occurred. For instance vertical is very simple (fast) and can be used for airborne data for instance etc. The horizontal mode is more unlikely, but it can be used for cliffs for instance (and I'm not even sure :D). And the 3D mode is more generic.

2) Core points are points on which the values will be computed and then transferred to their nearest neighbors. The idea is to compute the distances on less points in order to go faster and by assuming that values don't change so fast locally. Of course, once you have the right parameters and you assessed that on a few points, you can restart the computation with all the points.

3) The 'guess params' button computes the parameters by randomly selecting points and their neighbors. So each time the results might change due to this random selection, especially if the density of the cloud is not regular. To choose the right values, you can try to guess a kind of average tendency? Or think about what these values really mean (but you'll need to read the article carefully).

4) For volume calculation, are you referring to the '2.5D Volume calculation' tool? Then the parameters highly depends on your data (do you have a screenshot at least?)

5) Once again, you can use 100% overlap in ICP if the coverage / overlap of both clouds is exactly the same (at least all the points of the aligned cloud have an equivalent in the reference cloud). Generally it's not the case (especially since you have generally differences). And you can go as low as 10% if both clouds are already well registered (and if you really need it of course, if there are a lot of differences).

6) For adjust scale, it's only necessary if you have some doubts about the scale of the aligned cloud (if it's a photogrammetry cloud for instance, and you trust the reference more).

[Adam]

Thanks for your reply :)

Yes, I did read them.

What is 3D normal? as far as I know, the options are: default, multi-scale-vertical, and horizontal. I normally use vertical. My data is from Triangulation based 3D laser scanner. I have scanned some concrete material surface.

Yes, for volume its 2.5D volume calculation to look at the volume losses.
When I use %100 overlap and then calculate the volume I get way different results from when I chose %80 overlap or less. SO it confuses me as I want to know the materials losses.
when you mean exactly the same, does this apply for the shape of the material? because when you use a laser scanner there will be small differences and the concrete material surface might get eroded. so I'm trying to measure the erosion rate.

Here is a screenshot:

thanks

[daniel]

"Default" is 3D. That means that the normal is point in any direction (it's actually the normal to the surface, so the direction will depend on the surface).

And for the overlap, it seems you do have nearly 100% here. But also the two clouds seems so close that you must mainly measure noise! Hence the sensitivity to the parameters.

[Adam]

Thanks for the reply.
So in my case, should I use vertical? and does vertical tells me?
I have a doubt that I should be using 100% overlap because sometimes the shape and the points are not exactly the same + some concrete surface has been weathered/Deteriorated. So I'm using 80% overlap which I wanted to confirm with you if this is a reasonable overlap percentage or not, and what are the results of 80% overlap means.

do you mean by the two clouds seems close as they are almost the same? they are the same surface: one is the reference (unweathered/undeteriorated) point clouds and the compared one is the weathered/ Deteriorated.

How to measure the noise?

Regards