CloudCompare forum

Hey, i am quiet new to Cloud Compare and have a question regarding the optimization of point clouds.

We Scan mainly Buildings with the VLX Scanner, and have mostly the Problem of Noise behind of severel Windows. Especially houses with window facades are terrible.

I search for an quiet automatic solution. Atm we use Scene for cutting off all this stuff and we want to automize this step as far as possible

I found the tool SOR which is nice, but the noise points behind the Windows are sometimes so close together that i cant differ those from the points i want to keep.

Then i found the Plugin Canupo and tried to Classify some stuff but to be honest i didnt manage yet to look at some tutorials here - does it do its magic in this case?


I've added some Screenshots to show my problem i think there is a simplier solution than manually cutting all the sides off.

Hum, sadly no, I don't think Canupo will help you with that... that's some ugly noise!

Hi Timo,

I saw this video on you tube yesterday that was able to eliminate some noise by modifying the scalar field.
https://youtu.be/jH6njZGpZPE

Perhaps this may work in your case as well.

Thank you both for the answer.
What a pity but i already thought that canupo is not for this purpose...and yeah this noise drives me crazy ^^

@cdordoni - i just tried this...its a great feature, but still skips a lot of the noise. Also we have in this case some blue colored Walls inside which he also recognize of course on the B channel.

Its not just noise where some Sky Segmentation would help, its mirroring the room on the inside to the outside...which makes it difficult to select the color..

in the end i think the best solution is in the optimization of the Scanning Device or better how we scan in "heavy windowed" areas.

Processing with high confidence in Ivion might help clean up some of the noise, as the filtering is based on the distance from the scanner but you have likely tried that already.

Does going trough the glass count a return? if so, you could expose the retunr numbers and filter anything with more than 1