CloudCompare forum

Hello,

within my master thesis I am trying to classify soil surface traits on very small plots (55*55cm). With CANUPO, I am trying to manage to recognize objects like leaves, stones, gravel etc. not bigger than 15cm. During the classification process I managed to reach good parameters (ba=0,99 ; fischer discriminant=13), but when I apply the classification on the point cloud the single objects always have too big contour effects for my purpose. The pointclouds are quite dense (up to 100000 points per square) and are created using photogrammetry and PhotoModeler.

Any suggestions or ideas about how I could solve the problem? Or did anyone had the same issues during some projects?

Thank you,

lorenzo

What's the biggest scale compared to your cloud width (55 cm)?

The biggest scale I applied went from 2 to 25cm. By reducing the biggest scale to 20 or 15 cm the precision of the categorizer was not always as good and the point clouds more difficult to differ for CANUPO.