CloudCompare forum

Hi Daniel

I was thinking of a simplified method of bare earth extraction that uses the octree structure:
1) define a level of octree subdivision
2) divide the point cloud into columns of octree cells and choose the lowest point of the lowest octree cell of each column
3) create a new point cloud with points selected in step 2
it should be a quick and easy way to remove sparse vegetation, people, cars and vertical or overhanging objects from scans

Thanks
Alberto

Hi,

I think the method already exists: it's in "Tools > Projection > Height grid generation".

This method let's you decide of a regular grid step (~ level of subdivision), an orientation (X, Y or Z) and then projects all points in 2D along this orientation, keeping only one point per grid cell. You can eventually choose which point is kept each time (max, mean or min altitude - the latter one will correspond to your proposition).

I believe some users already use it to extract bare earth from aerial lidar data with it.

Thanks for the suggestion anyway (and tell me if I missed something!).

yes, something like this

the heigh grid generator forces the creation of a regular grid, can you only sample the existing points?

Good idea. We should add an option to do this with the Height Grid Generation tool. I'll add this to the TODO list.

thanks Daniel

Ok that's done! It will be available in the next release (and it's already available on the github's trunk for those who compile the project).

thanks again, I'll wait for the next release

Does the height grid generation tool still exist? I do not see it under tools > projections.

Thanks

Yes, it's now called "rasterize" and it is much more evolved than the Height grid generation.