Support Website Contact Support Blog

Reoving Trees from the DSM

I am trying to improve the DSM by removing trees from being used in final DSM. I am not able to use the DTM as I find that it introduces huge errors in its smoothing algorithm (like 8 meters). The DSM looks great, but I need a clean surface (no trees are significant shrubs) as I am trying to create a surface model that will be used for mine planning. The area covers about 1,000 hectares and is mostly good for the surface (low grass, gravel roads) but on north facing slopes I have a significant amount of trees and taller shrubs. But, I still have probably 3000 trees that show up in the DSM. I have Global Mapper and thought about tracing contours that reliable (Taking out the tree influence), but I see this taking a good solid week. Any easy way to remove trees in a timely way? I have attached some images showing DSM, DTM, Ortho.

If anyone has other ideas, including using 3rd party software, it would be greatly appreciated.

Are there any GCP’s in the flight?

By the sounds of it - my position is the same as yours - Mining Engineer.
If your doing open pit mining and uploading the surfaces to your heavy equipment for grade control, I think the easiest way would by to just use the DTM and manually correct any noticeable errors.
Get your surveyors out of the lunch room ( lol ) and send them the the tree location to take some topo points every 15-20 meters and manually create that surface.

If its planning for underground, your DTM is going to be a +/- again get the surveyors out there to get those topo points.
But honestly im willing to bet your DTM is going to be fairly accurate as is.
Plus its the planning phase, if anyone notices down the road that its out of wack, blame the discrepancy on overburden lol.

1 Like

Dan, Did you try editng the point cloud (https://support.pix4d.com/hc/en-us/articles/202560499-How-to-edit-the-point-cloud-in-the-rayCloud)? You can classify the point cloud, put the trees in the disabled group, save the project and then process step 3. Also, have a look at this community post: Remove vegetation from orthomosaic, might be helpful.

Momtanu, I did try and reclassify the vegetation which worked quite well, but the majority of the areas remove large swaths not leaving enough ground points. In other words, when I go and create a new DSM, it leaves large flat areas that don’t actually reflect the ground. Can I draw 3dpoly lines in PIX4d that would snap to points I want to use (actual ground / not vegetation ) and then create a new DSM using those lines? Thanks

Dan, unfortunately, that cannot be done. I understand what you said, you must have canopy covering the ground so the points of the ground that can be seen are really less. It would indeed be interesting if we had something like that. Did you find any other tool?

Depending on your terrain you might be able to use this workflow.

1.http://www.cloudcompare.org/doc/wiki/index.php?title=CSF_(plugin)
2.https://www.cloudcompare.org/doc/wiki/index.php?title=Rasterize#Empty_cells
3.Import into Pix4D as a cloud to build a DSM (with the trees removed).

Please note that the CSF might work more effectively if the terrain is binned into similar topographical levels. Additionally, the limitation of the interpolated rasterization is that it limited to the convex hull of the non-empty cells. If you would like more information please feel free to email me.

jason@geodronesurvey.com

Thanks Momtanu and Jason, I am still working with various options. I will update as I go forward!

I do have many GCP’s.

I failed to previouswly attach the images showing the issues that I find with the DTM smoothing it does. The only solution I see is to use the DSM, but still need to remove the vegitation.

1st image is the DSM. The green contours are made from using the DSM. The brownish orange ones are from the DTM.

2nd image is the DTM, same contours.

3rd image is to show how large of a discrepancy you can have between the DSM and the DTM.

4th shows duscrepancy again bt the DTM smoothing out areas. This adds up to a tremendous amout of volume over even rather small areas.


![clip2|690x354]