Project

General

Profile

How to obtain clear cluste of stem? and what could be the problem of "cutted trees"?

Added by Torresan Chiara about 3 years ago

Hello,
following your tutorial “Plot processing Step 2 Preparing Vegetation” from the step “Euclidean clustering denoising (0.10, 1500, 1, 1)” I got the result in the image. As you can see, I have not clear stems and there is still lot of noise due to the trees that are in the ground or due to other vegetation in the ground. How to solve this issue? I mean how to obtain clearer stem?
And then, from the step “Smooth extraction of a point slice parallel to the DTM (4.95, 100)” I got the result in the image. As you can see there are trees that haven’t the lower part of the stem. What could be the problem?
And two more question.
I upload the image of a cherri tree obtained from the QSM with allometric correction. As you can see, the stem is interrupted at a certain height (there is not cilinder). I had not problems of occlusion in my area. What could be?
And in some case from the application of QSM with allometric correction I obtained a cluster that has lot of tree (and so it results as a single tree), see the image upload. How to deal with this?

Thank you.
Chiara

Results_from_euclidean_clustering_denoising.png (241 KB) Results_from_euclidean_clustering_denoising.png Result from Euclidean clustering denoising
Largest_cluster_plus_slice_above_DTM.png (76.2 KB) Largest_cluster_plus_slice_above_DTM.png Result from Smooth extraction of a point slice parallel to the DTM
Cherri_tree.png (20.2 KB) Cherri_tree.png Prunus avium
Tree.png (37.4 KB) Tree.png Single cluster with lot of trees
1631
1632
1633
1634

Replies (3)

RE: How to obtain clear cluste of stem? and what could be the problem of "cutted trees"? - Added by Hackenberg Jan about 3 years ago

Hi

a) euclidean clustering: your screen looks like your cloud is a little bit better quality. That means you can set the clustering threshold from 0.1 to 0.05 or even lower. Try out. With this step you will always have still branches connected to your stem though. It is only to get rid of artefacts.
b) No lower part: you most likely have lost it during the previous steps, a.k.a. during Euclidean clustering denoising. Try to set the minimum points from 1500 to 500. Check if this changes the results
c) Looks like a screen from cloud compare. I have up to 7 folders where I store ply files. I kicked a lot of them out for my next release. Is the folder with “color” better? I assume you have cylinders, but they might not show in this ply file. Try the folder with “color” named.
d) see answer a) –> lower threshold in clustering –> more clusters. If this does not help, try out cutting higher above ground in “Preparing the Vegetation” steps.

Jan

RE: How to obtain clear cluste of stem? and what could be the problem of "cutted trees"? - Added by Torresan Chiara about 3 years ago

Hello,
yes, after different trying, I understood that “Euclidean clustering denoising” is the critical step to eliminate small stem in the ground. In particular the clustering distance, the number of points per cluster and the percentage have to be setted according to the specific condition of the plot. And yes again, lost part is related to the setting of “Euclidean clustering denoising” step.
I have not the “color” folder from the export. The folder I obtain are: allom, bad, detection, good, improvement, leave, no-leave, stem. The image uploaded was from the “good” folder.
Now have another question: as you can see from the image attached, I have many stems with a cylinder with a large radius. In this case, what could be the reason?
Thank you.
Cheers, Chiara

RE: How to obtain clear cluste of stem? and what could be the problem of "cutted trees"? - Added by Hackenberg Jan about 3 years ago

Hi Chiara

the folder with “color” is already included since more than one month. I highly recommend you (if not shortly before producing final results to be submitted) testing out newer versions. It is up to you and also understandable if you do not want to change the version much.

To your question:
always look at your input data (the cloud from the last step before modelling). I expect that there is some ground left around the stem leading to the minor errors. I anyway recommend running an advanced stem taper in R over the results to improve such errors. I heared of a package called TapeR, never tested it. You will find resources for that in traditional forestry for sure.

Jan

    (1-3/3)