Rough and Bumpy DTM for vegetated/forest area

mrshaz

Hi, 

I have lidar data (UAV-Lidar) for an area consist of roads, buildings, and ground covered by shrub and trees. I did the auto-ground classification and then continue with creating DTM. However, the terrain model for area that covered by shrub/trees  looks too rough and bumpy, as in the picture. I've tried and error here and there like reclassify the ground using different parameter, reduce/increased the bins size etc. but nothing seems to work. I've also tried using the painting tools but smoothing them just doesn't seems right to me (the resolution for that area looks bad) and the bumpy is still there.

Anyone have an experience like this? or any suggestion on how to get smooth and clean terrain surface for area covered by trees/vegetation...

Thank you.

XionicFire

Ditto on this problem

schsan

Hello,

I have the same problem. My workflow is to reclassify the vegetated area with the path profile tool. But that is very time consuming and not a very good solution. I do this because I cannot figure out another workflow...

bmg_mike

When you created the DTM, did you use the 'Filter Lidar' button on the dialog to restrict the gridding to only the ground points? If not, you would just be doing a 'Bin Minimum' grid on all Lidar return types. You would likely need to use a larger gap fill to fill in holes if there aren't many ground points being found under the low vegetation.

Thanks,

Mike

Global Mapper Guru

XionicFire

cant speak for OP but at least for me, yes i turned off everything but ground, then ran the DTM for ground

bmg_mike

It must be that the rough low vegetation was being recognized as ground as well, or at least some points inside that vegetation.

Our new Lidar Spectral Partioning tool can often do a better job at finding ground points in some complex situations like this. For example, the tool has been run on the data set below and assigned different clusters of points to a unique segment ID. As you can see, all of the ground points are in a single segment. You can then use the Select by Segment ID option and select all of the ground segment at once and assign them to the ground type.

We plan on updating our ground auto-classification tool in the future to allow using this more advanced (but slower) method of identifying ground points. Our building and high vegetation automatic classification can already make use of this method or the older gridded method.

Thanks,

Mike

Global Mapper Guru

schsan

Hello Mike,

yes, i always disable all other classes. With which settings have you run the segmentation tool at your example image?

Greetings schsan

bmg_mike

I don't have the exact settings used for that display, but something like below gives me 2 large segments for all of the ground points. Those 2 segments could easily be selected and re-classified:

Thanks,

Mike

Global Mapper Guru

khurranakabeta

I have the same problem. My workflow is to reclassify the vegetated area with the path profile tool. get-mobdro.com

DerrickW

White box tools has some great functions for this, and a standalone gui. Look for "remove vegetation residuals".

https://www.whiteboxgeo.com/manual/wbt_book/available_tools/lidar_tools.html

dsmithdsmapping09

Im having the same issue. Cannot figure out what the best settings are to exclude points above the ground. Would appreciate some pointers!

BMG_Amanda

Hi dsmithdsmapping09,

The setting for excluding points above ground is "Max Deviation", also called "Minimum Height Departure from the Local Mean" in v24.1 and older. This tells GM how thick your ground points are supposed to be. Lowering this value can help to exclude vegetation, etc.

Another option is to use segmentation as Mike mentioned above, taking advantage of the Max Curvature threshold. If you are using v25, you can achieve the same thing with the Max Likelihood option for automatic ground classification.

Happy Mapping

Amanda

dsmithdsmapping09

Thank you Amanda!

dsmithdsmapping09

Was able to produce much better results, however my dtm is still pretty rough and bumpy. Are there any tutorials through global mapper that discusses this particular issue? I can run this in Agisoft and generate really good results but I want to use Global Mapper for the myriad of tools and I believe GM can ultimately produce a better end product. The point cloud Im working with is fairly wooded. Adjusting the max deviation helped but there is a fine line because it starts to remove features I want (like roads).

BMG_Amanda

The answer to this is based on your data. Go ahead and send it over to the tech team (geohelp@bluemarblegeo.com) and they will assess settings options for you. Feel free to link this thread, and if possible include the example from agisoft so we know exactly what you are looking for.

Thank you

Amanda