Project

General

Profile

Forums » Forum Public »

SimpleTree Segmentation into trees

Added by Bott Frederick almost 4 years ago

Hello Jan, Hello CompuTree Community,

I’m currently working with CompuTree and SimpleTree for my master thesis. So at first I want to thank you for it, for the good Documentation and all the help you give in this forum.

I have a high resolution mobile LiDAR Data set in which I am segmenting hundreds of trees. The segmentation of the trees with the tools Euclidean clustering operation (ST), and segmentation into trees (ST) is actually working pretty well ((even though some stems have some gaps because they were concealed of leafes) - Screenshot attached) and since the update to Computree 5 the las. Export into TreeClouds is also working pretty good.

My simple question about the segmentation is: How is the tool “Segmentation into trees” working?
I read all your papers about SimpleTree and tried to looked into the script of this step “SegmentationAll” but couldn’t actually find it. Is it already part of the Spherefollowing Methode? Or is this is already part of the QSM modelling of the trees (I’m just need to segment the trees)

Best Frederick Bott

brief workflow:
-Statistical outlier Filtering
-Voxelization
-Classify ground points
-Create DTM
-Smooth Extraction of point slice parallel to the DTM (stems and crowns)
-Statistical outlier Filtering
-Euclidean clustering denoising
-Euclidean clustering operation (of stems)
-Segmentation into tree clouds

SegmentierteBaeume.PNG (97.7 KB) SegmentierteBaeume.PNG
2625

Replies (5)

RE: SimpleTree Segmentation into trees - Added by Hackenberg Jan almost 4 years ago

Hi Frederick

in short:
  • You import both the plot scenery and a vector of seed clusters. Each seed clusters’ point gets the unique cluster id and an initial distance of 0.
  • Then a competitive Dijkstra’s algorithm is run. Each point is connected to one of the clusters and receives that according clusters id. A graph is build by connecting neighboring points and there has to be a path between a connected point and the seed cluster. There is multiple paths between a target point and any one of the seed points. The path with the smallest distance is chosen.
That is a bit simplification but it explains everything in its behavior.

Hope this helps
Jan

RE: SimpleTree Segmentation into trees - Added by Jurjevic Luka over 3 years ago

Hello everybody,

I don’t want to open a new topic so I will write here as it is kid of related.

So my question is is there any way to export segmented trees (using Simpletree) as individual point clouds (other than manually selecting and exporting each tree)?
So basically export each cluster as separate point cloud?

I have to do it manually so far - export to ascii with point ID, import to matlab and run the script to export based on the point ID etc, and then proceed with cloud compare analysis. Is there some shortcut to do that?

Best regards,
Luka

RE: SimpleTree Segmentation into trees - Added by Piboule Alexandre over 3 years ago

Hi,

You can use export/point exports steps, using “export multiple files” option. (v5 of Computree here http://rdinnovation.onf.fr/projects/computree-beta-version-base/files)

Best regards.
Alexandre.

RE: SimpleTree Segmentation into trees - Added by Jurjevic Luka over 3 years ago

Thanks Alexandre! I was using the v4 Computree. Really nice feature, makes it so much easier for me!
Nice piece of software, keep up the good work

Luka

RE: SimpleTree Segmentation into trees - Added by V Matti about 2 years ago

Can you clarify your workflow:

-Statistical outlier Filtering
-Voxelization
-Classify ground points
-Create DTM
-Smooth Extraction of point slice parallel to the DTM (stems and crowns)
-Statistical outlier Filtering
-Euclidean clustering denoising
-Euclidean clustering operation (of stems)
-Segmentation into tree clouds

I tried running it and it runs into odd non-informative results from step 4 onwards.

I don’t understand:

Why DTM and not DSM/CHM?
What does step 5 do really? It results in odd point structures that I don’t understand.
How are steps 7-9 supposed to work? What kind of cloud do they expect as input?

    (1-5/5)