Simpletree - QSM advanced for plot crashes
I am working with the latest release of computree (V5) to generate the QSM of the trees in a 0.25 ha plot. My input is a las file of the pointcloud of these trees (15 trees in total). However after completing the whole workflow and successfully segmenting my trees, the final step (QSM - advanced for plot), crashes immediately.
From the tips I have read in the forum, I have tried several things to solve this, including: doing it for a single tree and translating the cloud to (0,0,0) to compensate for the fact that the cloud is georeferenced. However, still this step always crashes.
Does anyone have any experience of this and know what the error could be?
Thanks in advance for your help,
since you didn’t post your entire workflow its hard to tell, where the problem occurs.
But maybe there are points of the ground remaining after your segmentation? This caused some crashes on my machine as well. Did you filter your cloud in any way before processing (Statistical outlier removal e.g. with 6-8 neighbours and a STDEV of 2 -removes quite a bit of noise, but as well higher branches with less points might be deleted - use e.g. CloudCompare for that).
I would imagine, that Jan (developer of simpleTree) is not eager to read the following, but if you are able to get your hands on a Matlab license (most university have them) you can use treeQSM (search for a GitHub link). Its faster and I’d say a bit more accurate. I could send you a ready script for it and some explanations.
But if you would post your steps here a more detailed help on the crash could be possible as well.
Thank you for getting back to me! I have now attached my workflow here for you to look at. I would be very interested to hear any ideas you have about what may be wrong.
Also, I have tried using several filters within my workflow includung the Statistical Outlier removal, to remove noise (however the one I have attached is without this). In my experience it didn’t seem to change the fact that the QSM step kept failing.
I would be very interested to try treeQSM within Matlab. Any script or additional explanation you could send me would be very much appreciated.
sure no problem
Since I’m still a student I actually do have some exams until next Tuesday. Only after that I do have the time to reply in full length to you.
In addition. I’m not quite sure if we should discuss treeQSM in this forum... what do you think? - Sure, maybe other have the same problem and are eager to read about this as well..
So I’ll work with your point cloud on my machine in the lab next Wednesday. I hope you can wait until then.
You could start reading the manual of treeQSM, maybe you can figure some stuff out on your own und I only fill in the gaps.
Thanks for your help. I have sent you an email.
I’m a little late to the conversation, but have experienced similar issues in the past with SimpleTree. You went over a few of the main things that will lead to crashes.
In the LAS file you included it appears you have georeferenced coordinates, which need to be centered in order for the modeling to work. I always keep my point clouds locally referenced and then georeference the products or measurements from TLS with plot center.
The next thing I noticed was that your point cloud looked pretty sparse. If that the full quality version? If so, that could be the issue.
Finally, a while back I tried to implement my QSM script in Computree using LAS files (for compression, etc.) but kept having issues with crashing. Before, I always used the XYB format as this seemed to be Jan’s preferred format, but LAS seemed to make more sense. I took the exact same script and changed the input format and - voila! - it worked without crashing. So you may give that a shot before committing to a MatLab workflow (even though treeQSM works great!).
so I found out that your final segmented trees clusters did contain some not really good segmented trees. I attached two pictures which show exemplary what I mean by this. In addition, attached picture three has some ground point as well in it. So this could be the reason why the plot is not working with simpletree.
I attached a workflow which you can modify to import .las clouds or you use the second script (las_to_xyb, just right click on’Create a list of readers and then choose 'Config.parameters and browse to the file or multiple files you want to convert, do the same with the 'Points,XYB’ step to choose a target destination - adapt steps to rename the files correctly) to convert an .las file to an .xyb file and then use the first script (xyb_to_trees).
The workflow 'xyb_to_trees’ I is designed to let you manually adjust the segmented trees, and if needed reassign points to a different tree. This assures a quite precise pointcloud of every tree. See the manual I attached (I couldn’t find the original at this point on the website anymore, only the french version) to see what which step does and what you can do in the manual segmentation mode.
You then can add at a last step the center point cloud tool and the simpletree plugin to the workflow.
This approach is unfortunately no longer fully automated, bit it should be more precise in regard of which points are designated to which tree. This is a huge advantage in my opinion, well unless you have a heap of such clouds...
I attached as well my final treeQSM script. But to use this, you have to have single point clouds in an .xyz format (.txt / or .asc) for each tree.
Sorry for the late reply. I read the discussion and decided to try out modelling on the SimpleForest plugin, which I developed behind the scenes the last years.
- I took the xyb version provided by Tobias and imported it in a script used for high resoluted clouds (Bertts.zip - above post). The script did not crash, but had to be adapted.
- I was able to fully automatically segment the cloud without interaction
- The total script is also quite fast - I think less than 5 minutes of compuation, but this is depending on the hardware also . My computer is already 7 years old.
- I produced in two ways QSMs, see the better results here:
The QSMs and the cloud:
I attached the script to this post.
Not sure about las problem stated by Tobias. You can try out with las but no garantee. Otherwise SimpleForest should be able to handle geocoordinates, each step (hopefully each step) should transform before computing and transform results after compuation. Well I should have double checked before stating this, but I tested few months ago and then it was fine.
You can download latest windows release of CT with my new plugin here:
I will later this day upload tutorials on my youtube channel:
I need few hours still to produce the last one.
p.s. I cannot support my SimpleTree plugin anymore.
I would also like to give you information that if you modify parameters for this cloud after one compuation, CT crashes. This behaviour does not occur on the last steps after having the segmented clouds I think. As you can compute super fast until here it is not a big problem.
This did not happen with similar scripts on other clouds. I could always recompute whereever I am in my script pipeline.
I tried to contact you 1/2 weeks ago. Is your email still valid?