Underwater Sonar data

[multibeam]

Hi there,
What do guys think of this Poisson surface of an underwater Quay wall?

QuayWall.JPG (224.15 KiB) Viewed 2383 times

You can see the effect of different passes with the sonar yield slightly different positioning of the data points, this is due to a combination of small errors within the: GPS positioning, Motion sensor correction, differing water density etc.. and it makes the Poisson surface look messy/inaccurate.

My question is this, can these different sets of data be aligned within Cloudcompare? or should I just choose a lower density surface construction?


Cheers, any help would be much appreciated!

[daniel]

How many different sets have you? If it's 2 or 3 then you should indeed be able to improve their relative registration (if the overlap and density is sufficient of course).

Otherwise, level 15 for Poisson is huge (you can reduce this a lot - e.g. 10 or 11 - without changing the output quality much). And computing smoother normals (with the Quadric mode and a large radius) can help in your case, but only a little.

And to smooth the result you can indeed reduce the Poisson depth, but you can also use the 'Edit > Mesh > Smooth' algorithm.

[multibeam]

Thankyou, I will try this out!

Yes 2-3 sets usually. For the relative registration, is it 'point pairs picking' ? I never had much luck with this before.. is there a trick to it?

I only used 15 Octree depth because I am using a powerful (x99 6-core) PC.. and I thought the higher this number the better, which is obviously bit naive of me!

[daniel]

No you should try the automatic registration (ICP) with this kind of clouds. Even though they seem very flat... can you show me each cloud separately maybe?

[multibeam]

Yes, should I send you the data or just take a screenshot of each cloud?

[multibeam]

Heres a quick screenshot anyway:

Each set.JPG (163.44 KiB) Viewed 2378 times

[daniel]

You can send me the data at 'cloudcompare [at] danielgm.net'

It's always a bit dangerous to register very flat clouds this way... I'll see if it's possible.

[multibeam]

Thanks Daniel, I will send a the data!,