Point Cloud Tree Data

[townofmountainvillage]

Hello!

The Town I work for has recently received some pretty spectacular LiDAR point cloud data in the .las format as well as some other data products. We are wanting to use this data to analyze our forests and infrastructure. One of the goals is to get these large datasets into working condition and make them more usable. The point cloud data is enormous and not feasible in its current state. Is there a way to turn the vegetation into smoothed out 3-D features? I know this is possible for buildings based on this link: https://learn.arcgis.com/en/projects/extract-3d-buildings-from-lidar-data/ but I am trying to rework this procedure for our needs and struggling. The trees don't have to be perfect shapes (a cone is sufficient) but how would we go about doing that? We want the density of the forests represented accurately as well as the height/width of the trees. Has anyone done something similar to this?

[David_McRitchie]

Interesting question! I think there will probably be numerous ways to do this but it might be worth using Deep Learning for part of this.

The Living Atlas has a number of Deep Learning Models that are already trained and can be used against your own data, such as this one here for tree classification. The Item detail page provides a guide for using this.

This will get the classification part of the problem sorted, as for having these points turned into 3D symbols with height and density being reflected it will be more difficult. If it was me I might try use some aerial imagery with the following Deep Learning model to provide polygon boundaries of each tree, then do a count of the number of classified lidar points within each polygon and then use 3D tree symbols with varying sizes based on the number of classified lidar points counted in each polygon.

It probably isn't a perfect solution, and using two deep learning models will require a lot of processing time, but it might be a suitable method to get this working.

Hope that helps,

David

[EdwardBayes]

Hi David,

These are really cool! Do you have any other tips on the best ESRI plugins to use for these purposes? Are these generally known as the best ones? 

Thanks,

Ed