M2C-M3C2. Some questions.

[flavia.sistilli]

HI Daniel,
first of al, good new year!
I was wondering if the link with the file, that I sent you, had arrived.

thanks
Flavia

[daniel]

Hi and happy new year to you as well!

No I haven't received your email... did you send it to "cloudcompare [at] danielgm.net" ? (replace [at] by @)

[flavia.sistilli]

Hi daniel,
I'm almost sure that the email was correct,
anyway I have just resend you the link.
I hope it reaches you!

Flavia

[daniel]

Yep, this time it worked!

[flavia.sistilli]

Good!
So, I wait your opinion!

thanks
Flavia

[daniel]

Ok, the issue seems to come from the way you created the mesh.

It seems that you have used the the 2D1/2 Delaunay triangulation method. The result of this algorithm can look quite good from above, but in fact it creates some wrong faces on truly 3D clouds or even on 2D1/2 clouds when there's big variations of height:

2D1/2 Delaunay triangulation issue

cc_2D_delaunay.jpg (70.53 KiB) Viewed 5691 times

Those wrong triangles can really jeopardize the C2M distance. To properly mesh this kind of cloud you have to set a maximum edge length (something like 0.05 on your example) so as to avoid creating those big triangles on the sides:

2D1/2 Delaunay triangulation (best LS plane) with max edge length = 0.05

cc_2D_delaunay_max_edge_length.jpg (114.74 KiB) Viewed 5691 times

With this mesh, the C2M distances computation gave me the following result:

Cloud to mesh distance

cc_c2m_distance.jpg (158.85 KiB) Viewed 5691 times

Properly meshing the vegetation is almost impossible here so you still get strange 'inverted' patch near the vegetation or holes.

Maybe a good idea would be to remove the vegetation first (with qCanupo for instance ;). And of course M3C2 is clearly superior.

[flavia.sistilli]

Hi Daniel ,
I have used Delanaunay 2D (XY plane) for mesh,
but also using other algorithms, the result is the same.

Now,I'll try the different solution, to use Canupo, to set a maximum edge length..
Actually using M3C2 there isn't this problem,
even if the end result is not what I would, graphically.
Probably it will be the most convenient solution.

thanks very much for your time!

Flavia

[daniel]

What is exactly the issue with M3C2's output?

[Dimitri]

Using M3C2, if you select the option to output the standard deviation, you'll be able to remove from the results the area where vegetation affected the calculation (high standard deviation). Unless you manage to remove the vegetation with CANUPO (this implies creating your own classifier), meshing your data is likely going to give poorer results.