Measurement plane, a small cloud of points - Ransac Error

[Mariush]

Hi, I am Mariusz a student of The University of Bielsko-Biala from Poland

I need a help in basically operation in Cloud Compare.

I make nowadays lecture about Ransac algorithm, automated plane detection and measurement. I want to show How this Feature works in CloudCompare 2.7

I have a simple point cloud caught by Laser Displacement Sensor „Keyence” mounted on industrial robot.

So I have only co-ordinates of points (X,Y,Z) in .txt file. Size of cloud is only 25 points.

I would like to Cloud Compare detect best plane fitted to the measured points and designate equation of the plane for me.

I watched the tutorial on YT https://www.youtube.com/watch?v=uUBUh24hDQI and it seems to be simple. In the movie author only match the cloud and run Ransac Plugin with default settings. (I only change minimum support points per primitive to 10) When I do the same I mostly have error: „Segmentation Failed” , in some cases CC was crash.

The only difference that I notice is the point cloud in Tutorial is more complex. Is the Mesh object turn to cloud with 80 000 points,

So, my question is: What is the problem? Default settings is wrong to my object or cloud type is wrong for this case?

I have the same problem in 2.7 and 2.8 Beta version, I tried on Win 10 and 7 64bit version.

I found on forum that Cloud Compare not support small clouds very well, so maybe that's the reason?

Small Clouds: I consider a cloud to be 2 or more data points. Small clouds don't behave well in Cloud compare. RANSAC, Connected Components and some other functions either crash the app or hang or go off in an endless loop when this is the case.

by andrewo in "functionality enhancement suggestions" topic.

Best regards

[daniel]

Indeed, the issue might probably be the number of points. Technically, in this case it's probably not an issue with CloudCompare, but more probably with the 'Ransac Shape Detection' library that is used by the plugin (but one thing is clear: both this library and CC are definitely meant to be used with much more points ;)

If you want to fit a single plane on a set of points (whatever the size), then you can use the 'Tools > Fit > Plane' method (it uses the Least Squares fitting method). Ransac Shape Detection only makes sense if you have a much more complex structure with several planes. And you definitely need a lot more points.