M3C2

[Ax_Ahmed]

Hi Daniel,

I have 2 scans where I have used the M3C2, but the height difference looks a bit odd. Does it look right to you? I have attached the pictures.
I still have a problem with the scalar field as the samples come in a blue colour only.



Thanks

[daniel]

Maybe the easiest way for me to 'debug' your case would be to send me the clouds...

[Ax_Ahmed]

Hi,

Alright, thanks. I have just sent them to cloudcompare@danielgm.net. Please let me know if I'm doing any mistakes especially with the M3C2 setting as I keep trying a different setting. I'm keen on the local standard deviation STD1 and STD2 also to get as much as accurate results to present the roughness. I wanted to look at the height difference between the 2 clouds with numerical evidence. I have sent you 3 files please don't share them as these belong to my thesis and paper :)


Thanks

[Ax_Ahmed]

Hello,

I just would like to check if you have received my email?

Regards

[daniel]

Yes, but I didn't find the time to process it yet...

[daniel]

I did a quick ICP registration starting from the two meshes I found in the "Scan 1&2_Final rig" file.

Here is the result I get with the Cloud-to-mesh distance:

cc_c2m_dist.JPG (101.19 KiB) Viewed 4214 times

And here is the result I get with M3C2 (i.e. almost the same):

cc_m3c2_dist.JPG (105.25 KiB) Viewed 4214 times

I used 80% overlap for ICP. And a scale of 0.11 for M3C2 (with vertical normals).

[Ax_Ahmed]

Hi,

Thanks for getting back to me.
1- The 0.11 scale is for normals and projection? and why you have chosen this value? is it ok to use this value for the other samples?how about the depth?
2- I never get a map like that with M3C2! how did you get it? all that I get is looking like the one uploaded below.
3- Overall with the files I have sent. Am I on the right track? Could you look at them when you have time and send them back to my email if that's alright


Regards

[daniel]

1. I used the 'guess params' buttons :D. I just specified the value I used so that you could try to reproduce it on your side. However I think the issue on your side might be the registration, and not M3C2.

Note that there's no 'normals' radius necessary here as I used purely 'Vertical' normals. And the depth depends on how far are the clouds, and if you are really interested in getting very big distances or not (in which case you can accelerate the computations by limiting the depth of the search cylinders).

2. For the map, at least you should use a much more dense core points cloud (reduce the sampling distance, or use all the points of the original cloud). And then I only played with the scalar field saturation parameter (I put the saturation on 0.5 in my examples, to get the default color ramp to span between -0.5 and +0.5).

3. I'll try...

[Ax_Ahmed]

Hi,

I tried my best for the registration and I used both Point pairs picking and ICP which I think it works fine.
For the map how to reduce the sampling limit or use all the points?
Where is the saturation parameter?
Is the map above for M3C2 only? not cloud 1 or 2? I never get it the way you did it.
Please look at the other files :)

Merci!

[Ax_Ahmed]

Hi Daniel,


I haven't heard back from you regarding the issue below and my files that I have sent a while ago.

Regarding tp tools>fir>plane:
Does this work for concrete cubes? and the way it works is by selecting both clouds (at the same time) then using the tool. It will create a matrix of each cloud. I should only pick one matrix for transformations (any particular clouds 1 or 2?)
The image's colour doesn't look right.

Thanks