Cloudcompare M3C2 vs C2M signed distance outputs (reference cloud problem!)

[soheilsabz]

I have two mode which using the cloud compare for comparing those together, i used the ordinary distance compare method ( C2M signed method) with this parameter as you can see here:



and for more details i tried also the M3C2 method by this parameters:



but there is difference in output as you can see:




So i think it comes form the M3c2 setting (which one is reference ,cloud #1 or cloud#2?!).

I have tried to give the reference stl file as cloud#1 and target to be compared as cloud#2 as at export part as projection core points on cloud#2 i have tried projection core points on cloud#1 but i have get the same output and it project the colored compare results only on the reference cloud which is Cloud#1 !.

so what is the reason for difference and why it only show the cored output on reference could?!

So i think it comes form the M3c2 setting (which one is reference ,cloud #1 or cloud#2?!).

I have tried to give the reference stl file as cloud#1 and target to be compared as cloud#2 as at export part as projection core points on cloud#2 i have tried projection core points on cloud#1 but i have get the same output and it project the colored compare results only on the reference cloud which is Cloud#1 !.

i have tried to use cloud#2 as reference and the **M3C2 parameters** are this kind now:


and the colored output shown on the target stl mode as you can see here:


, but the difference in output is remained again! you can see here:



so what is the reason for difference and why it only show the cored output on reference could?!

asked this Question also here:

https://engineering.stackexchange.com/q ... d-problem

Thanks.

[daniel]

So the two computation methods are definitely very different, and you have to take good care when setting the M3C2 parameters indeed.

Moreover M3C2 is not meant to compare meshes, but clouds. So you'd better convert your mesh(es) to dense clouds first (with 'Edit > Mesh > Sample points').

Then M3C2 looks for the nearest neighbors along the input normals. While the C2M distance just looks for the nearest points.

And last, you should not focus on the maximum distance which might only be set to a single point (= outlier) in the whole cloud. You'd better use an absolute color scale (see https://www.cloudcompare.org/doc/wiki/i ... es_Manager) and use it on both clouds to see how and where the distances actually are different. You could also compare the histograms.