Rayextract trees output individual tree point cloud and corresponding mesh files #33

woutervdbroeck · 2025-05-28T09:37:55Z

woutervdbroeck
May 28, 2025

Hi there,

Currently, the output of the rayextract trees command is the original forest point cloud segmented into individual trees (represented as different colors), a treefile.txt for cylinder representation, and a tree mesh file for the entire input. However, would it be possible to add an option to have the output as all individual trees? That is, for each tree the segmented pointcloud, the (single line) treefile, and corresponding mesh (e.g. tree_0_pointcloud.ply, tree_0_cylinder.txt, tree_0_mesh.ply, tree_1_ ...). My application requires manual quality control to check that each segmented tree pointcloud and corresponding mesh makes sense. Hence, having all trees output individually would be very useful. It may also be cleaner than having to split the segmented pointcloud based on color and matching all trees based on e.g. stem base coordinates or bounding box with their corresponding mesh/cylinder file.

Hope my question makes sense. Thanks in advance!

Cheers,
Wouter

thomas-lowe · 2025-05-28T11:42:15Z

thomas-lowe
May 28, 2025
Maintainer

Hi Wouter,
yes, this is possible. To cover more options like this there is the library treetools: https://github.com/csiro-robotics/treetools. If you used the docker image you should already have this helper library available.
To get separate tree files run treesplit cloud_trees.txt per-tree
To get a mesh for each of these files you will need to run treemesh cloud_trees_x.txt where x is the tree index. A bash script can run through the files, see for example the script in the raycloudtools scripts folder.
To get the individual point clouds from the segmented cloud run raysplit cloud_segmented.txt seg_colour

The suffix from the raysplit should match that in the treesplit because the colour scheme encodes the tree index so that's how it knows the suffix.

let us know if there's any problems you get.

0 replies

ZaneTCooper · 2025-05-29T12:27:06Z

ZaneTCooper
May 29, 2025

Hey Thomas,
Thanks for the quick response, it was already very helpful. We've been using the treetools library for a while but had missed the seg_colour option in raysplit - it should simplify things going forward. We took another look today and noticed a few things:

The indices of the raysplit segmented tree point clouds and the treesplit tree.txt files (and hence generated mesh files) look to be different for us. The segmented trees are starting at 0 and the tree.txt files are starting at 1.
raysplit is creating a point cloud with the index -1. We were hoping this would be all the remaining points that weren't used for any trees from that tile, but when loaded into cloud compare it looks like the below image. Is this an error in cloudcompare display or is something else going on here?

We are using the rayextract_trees_large.sh script to help us process hectare size point clouds. Because of the way that the index is assigned from the colour, using raycombine on the tiled point clouds followed by raysplit seg_colour to extract individually segmented trees will result in multiple trees being combined into each file (tree 1 from each tile, then tree 2, etc.). The solution is to run raysplit seg_colour on each tile individually. The names of individual trees will be a combination of the tile name and tree name, e.g. *0_0_segmented_1.ply, and the tree index will repeat for each tile. This can be fixed by renaming after the fact, which I might work into a batch script if I want to batch all of this at some point, but right now the raysplit -> raycombine method to tackle larger pointclouds wont yield unique tree IDs, unless you have another way we should approach it.

Thanks,
Zane & Wouter

1 reply

8000

thomas-lowe May 29, 2025
Maintainer

Hi Zane and Wouter,

thanks for letting me know, I have pushed the fix to treetools so that indices start at 0, to match the raysplit seg_colour.
something else is going on. When you use raysplit grid, it cuts rays along the grid planes leaving some rays with no end contact, these are marked with alpha=0 and visually with colour=0,0,0. These 'non-return' rays (like rays that go up to the sky) are not considered as surfaces in any raycloudtools operations and are useful to keep for operations like raytransients, but if you want to remove them you can use raysplit cloud.ply alpha 0. Ideally cloudcompare would render with alpha as transparency.
yes, you are quite right that raysplit grid complicates the raysplit seg_colour / treesplit per-tree combination. What you should do is per-grid tile x,y: raysplit cloud_x_y_segmented.ply seg_colour, and treesplit cloud_x_y_trees.txt per-tree, giving:
cloud_x_y_segmented_i.ply and cloud_x_y_trees_i.txt
The individual _trees.txt files should be in one of the output folders from the rayextract_trees_large.sh.
So they should all match, its just a combination of tile index and tree index. To make it a pure tree index you would have to do some renaming with python code I expect.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Rayextract trees output individual tree point cloud and corresponding mesh files #33

Uh oh!

{{title}}

Uh oh!

Replies: 2 comments 1 reply

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Select a reply

Uh oh!

Rayextract trees output individual tree point cloud and corresponding mesh files #33

Uh oh!

woutervdbroeck May 28, 2025

Replies: 2 comments · 1 reply

Uh oh!

Uh oh!

thomas-lowe May 28, 2025 Maintainer

Uh oh!

ZaneTCooper May 29, 2025

Uh oh!

thomas-lowe May 29, 2025 Maintainer

woutervdbroeck
May 28, 2025

Replies: 2 comments 1 reply

thomas-lowe
May 28, 2025
Maintainer

ZaneTCooper
May 29, 2025

thomas-lowe May 29, 2025
Maintainer