SOR with Distance, Select by Colour -> Automation/Customisation options?

[__G__]

Hi All,

Firstly, well done on a great program. I can see my self getting a lot of use out of CloudCompare in future.

I'd like to know what options we have for customisation - because I have some problems that I think may not be solvable with existing functionality.

Firstly - the SOR algorithm doesn't consider the distance from the scanner. I find that (in some case) when I use the SOR algorithm to denoise a scene, it just crops points that are a long way from the scanner - because they naturally get further apart (context: terrestrial scanner with ~70m range). I would like to be able to run the SOR algorithm with consideration given to the natrual tendancy for the points to spread out the further away from the scanner they are.
I think a work around is to divide the scan by different distances (say, 0 - 5m, 5m - 10m etc) and run the scan on each section, then recombine - but this would be made easier with some automation options (I see scripting is available - is this a suitable candidate for this task?)
A more robust (and more effort) solution would be to re-write the SOR algorithm factoring in a variable for the mean distance to neighbour based on the distance from scanner.

Secondly - I find that I often get 'sprays' of points on some slender objects that run overhead (pipes, beams etc) where the spray is a shade of blue (probably coloured by the sky). It would be great if I could select a range of colours so I can delete them from the scan. Currently, I believe the workaround is to add the colours to a scalar field, then trim them down one channel (R, G, B) at a time - this is difficult to visualise, and not very user friendly. Access to an API would be most helpful for this task I think.

Thanks again.

[daniel]

1) Sadly, I don't thing the current python scripting tools will let you change the behavior of an algorithm (but they might allow you to do this clustering approach you are referring to). The best way would be indeed to modify the C++ algorithm directly

2) You can now filter points based on their RGB values. See the new 'Colorimetric segmenter' plugin

[__G__]

Hi Daniel,

Thanks for the response.

1) - I see there is a Python library (CloudComPy) available - I'll explore this as an option for writing new algorithms.

2) - That seems to work... still not very user friendly (no way to preview), but it is much more simple than filtering by scalar fields.