CloudCompare forum

Hi

I’m having some trouble classifying a point cloud from a drone using the CANUPO plugin.

First of all, I get a quite good result following this video when I only have to classify ground and low vegetation: https://www.youtube.com/watch?v=8ryh_Qy ... ex=2&t=27s

My problem is that I have a point Clod where I want to make 4 classifications: Ground, Low vegetation, medium vegetation and high trees.

I’m using the latest version of CC where I can set the class of points inside the polyline using the # in the segment tool.

I create 4 classifications (Ground, Low vegetation, medium vegetation and high trees.) using the # in the segment tool.

After that I use the CANUPO Train Classifier, but here comes my problem: I can only choose two classes to make the classification (well that’s not quite right because I can choose all the classifications I have made, but the CANUPO plugin only handles to classes at the time). How do I make a classification of the point cloud which contains all 4 classes?

To summon up:

I can make a full classification when I only need to do 2 classifications in the point cloud.

I don’t know how to make the classification work when there is supposed to be 3, 4, 5 or more classifications in the point cloud.

Help much appreciated


Kind regards

Jacob

So the 'train' option of Canupo doesn't work like this. You have to provide it with individual clouds of each type (ground / low vegetation / etc.). And then you can tell optionally the plugin to flag the points with a particular value (class label).

So you have to first create a classifier for ground / non ground. Then Another classifier for low vegetation / non low vegetation, etc. And then you have to apply them in a cascading way.

And what you did with the segmentation tool will just let you create the 'training' clouds more easily.

Hi Daniel

It was the part of doing it in steps I missed. I will try that.

Thank you for answering

Have a nice weekend

Jacob

Hi Jacob, I am wondering if you ever got this to work and if you were somehow able to validate your classifications (ie a confusion matrix)?