Strange Segmentation Occurring
tool for manage and process TLS lidar data from forest environment
Brought to you by:
janekt,
m-a-r-t-in
Hi there,
I seem to be getting very weird segmentation. I am guessing maybe the axis are incorectly represented (see image).
Any ideas?
Here is another image.
HI
this problem occurs when wrong transformation matrix of project is used. Do
you know extent of your cloud? is yes please send me those coordinates and
I will send you values for transformation matrix that can be used.
Jan
st 19. 6. 2019 v 15:12 odesílatel noreply@sourceforge.net napsal:
Are the details that you need given in this image? (Bottom Left)
Is it normal for the point cloud to be separated into these parrallel lines?
I think the parallel lines are an artifact of previous sub-sampling
You Need values named As global shift in your screenshot. When creating
project, try to insert those values As new transformation matrix. Then
import cloud and sebe if the cloud looks normal or similar to present
state. If it not help use - oř + value in matrix...
Dne st 19. 6. 2019 19:08 uživatel noreply@sourceforge.net napsal:
Hi Janek,
Thanks for that, I managed to import without the parallel lines problem thanks to you advice on the tranfomation matrix. I used (x = -628400.00, y = -6891200.00, z = 0.00).
However, now I have problems with Terrain and Tree segmentation. See the picture
My Vegitation includes a lot of terrain and my trees are sementing as a clusters of Trees instead of individual trees. I assume this is becuse my parameters are incorrect; so I included the parameters that I used.
I'll keep troubleshooting nevertheless.
Hi, problem with terrain Is understory vegetation that can be solved
increasing to 12 or 15 cm instead of 10.
Better terrain segmentation can help also with vegetation segmentation. In
coniferous forest can help also decrease od values ito 3 cm Voxel size and
multiplicator to 2.
Jan
Dne út 25. 6. 2019 16:53 uživatel noreply@sourceforge.net napsal:
Thanks a Million Janek,
This is starting to look great. Most of my problems are now solved and I am continuing on. I'll likely have a few more questions in a week or two. Thank you kindly for the very useful support.
All the best.
Hi all.
I have succeded to segment 500+ trees. But I have a strange artifact. The bottom of each tree looses most of its points in the segmentation process. Is there a way I can get these back. I would like to have them so that I can look at the trees in other software too.
Hi can you Tell me what version you are using? And what parameters Are
set?
Jan
Dne čt 25. 7. 2019 14:23 uživatel noreply@sourceforge.net napsal:
Here are the parameters.
I am using Version 0.42
I just noticed you have released version 0.5
I will give that a try now.
I tried with Version 0.5 and it quit half way though.
I got 190 or 650(ish) trees.
I'll try again tomorrow.
Got 190 "of"^^ 650(ish) trees
(TYPO ABOVE)
Try to decrease number of elements when you set distance from the terrain
smaller. I would use distance to 0.5 or 1 m that can help with tree
segmentation.
Jan
čt 25. 7. 2019 v 21:09 odesílatel noreply@sourceforge.net napsal:
This time it got the whole tree though. I think what I did wrong was I had "distance from the terrain" set at 1.600
I think it should be more like 0.100
It seems that when I reduce the "Distance from Terrain" to around 0.500, then I only manage to segment 30% of trees in the point cloud.
I need to get the points around the bottom part of the tree, becuse the harvester cuts the tree at around 40cm or so, sometimes even lower.
I am a MSc student from Uppsala University. Janek, perhaps we can connect via email or skype. I also have some very interesting data. Inculding values from the harvester after harvest on this 1ha plot that I have terrestrial LiDAR for. I have sent you my personal details via the messaging portal at www.3Dforest.eu
Thanks a tonne for the assistance!
Hi
sorry for long answer.when you input smler distance to the terrain reduce
also nubber of elements to 20 maybe. that will select more trees. and try
to use bigger input distace. that can create bigger voxels and do not
exclude them. Hope this will help you.
if any problems contact us via web page contact form or by mail
j.trochta@gmail.com
Jan
so 27. 7. 2019 v 10:22 odesílatel noreply@sourceforge.net napsal: