Removing points from overlapping scans by specific scans

[Isaac Mui]

Hi all,

I'm fairly new to Cloud Compare and point clouds in general so apologies if this has been asked and answered.

We have been scanning with a RTC 360 and noticed due to slight variations in lighting/ exposure dependent on scan positions, it will create a "patchy" surface as the different scans are overlaid on top of each other please see attached images below.

I have tried decimating the points by space to reduce overlaps but in that case, it is still far from perfect and also lose detail which is undesirable. I have tried to experiment with a method in which two overlapping scans are compared with one being reference and the other(s) being compared and any points from the compared scans overlap with the referenced scans, they can be classified and thus removed. I was wondering if this methodology is appropriate to solve this issue and how to accomplish this.
Appreciate any help on this :)

p.s I understand that scanning in more controlled and optimal conditions will help reduce this issue but sometimes this is not achievable so would be good to come up with a workflow that can deal with problems like this.

[daniel]

If I understand well, your issue is only with the RGB colors of the 'mixed' points, isn't it?

[Isaac Mui]

Hi Daniel,

Yes that is correct. Scanning from different positions that are closer or further away from light sources alters the amount of light received from the onboard cameras that are used to colourise the point clouds resulting in this phenomenon. So my thoughts are is it possible to remove this overlap of points or is there perhaps a better solution to this issue.

[daniel]

Hum, not sure.
1) one idea if you have each scan as a separate cloud: you could compute the distances of one scan to the other, and remove the points that are too close to the scan with the 'good' colors

2) Maybe one other thing you could try is to:
- use the 2.11 version
- convert the RGB colors to scalar fields (3 separate SFs)
- on each SF, you can apply the 'spatial smoothing' (Edit > Scalar fields > Gaussian filter' - see https://www.cloudcompare.org/doc/wiki/i ... ian_filter)
- then merge the 3 fields back as one RGB field thanks to the new 'Edit > Colors > From Scalar fields' tool

The trick here will be to find the good radius for the smootihng filter so as to not smooth/blend the colors too much. That's quite experimental ;)

[Isaac Mui]

Hi Daniel,

Thanks for your reply. I had a try with the smoothing with scalar fields and it works in smoothing out the rough patchiness but i cant seem to convert it back into its true colours and is stuck with the "SF RGB colours"

12.PNG (292.67 KiB) Viewed 6461 times

This is the original colour.

13.PNG (1.16 MiB) Viewed 6461 times

Was wondering if i went wrong somewhere when converting the colour back into RGB.

Appreciate your help on this.

Thanks

[daniel]

Hum you're right, I don't fully grasp this new tool 'Edit > Colors > From Scalar fields' tool myself...

Can you maybe share the cloud with me to see what I manage to do? (admin [at] cloudcompare.org)

[Isaac Mui]

Cheers Daniel, I have sent it to you. Appreciate your help as always.

[Isaac Mui]

For anyone that is struggling with this issue, it is partially solved with Daniels recommendation above.
Converting RGB into scalar then individually smoothing each R,G and B value and reapplying the colour back into RGB through Colours -> From scalar fields. Alpha value is to be set to 255.
This is the original scan before this process:

15.PNG (643.5 KiB) Viewed 6344 times

And this is the scan after the process

14.PNG (557.3 KiB) Viewed 6344 times

Although not 100% clean, it does remove a majority of the granular effect and becomes more visually easier to read.
Hope this helps others too :)

[daniel]

Yes, thanks for the report!

And you'll need the 2.11.2 version to do that (as there was a bug in the 2.11.1). For those interested, the preview is here:
https://www.cloudcompare.org/release/Cl ... up_x64.exe

[mbradford]

And this is the scan after the process

Very new user here and having a blast with CloudCompare. This image here is a problem I'm seeing a lot! This before/after pic is incredible.

I see it has been a couple of years... Is this still the best way to handle the overlapped points that are very similar in color due to multiple angles of scans? I've been browsing the forums a lot and it seems like CloudCompare changes often; Just wanted to make sure there isn't a new feature/plugin/process that does this quicker.
Thanks!