Bumpy roads

Hello.

I have some vector data that represents the roads (urban and rural) of the area I'm working on. For the last few months I've been using Global Mapper to create DEM's and give elevation attributes to the vector data that my co-workers draw (in 2D). For the most part of the projects we've been working on, I've been able to get the results I need. However, for some reason, I've been getting a lot of problems in the current project.

For the last few months my workflow is the following:

  1. I classify and "clean" the point cloud (I generate DEM's as a trial and error to try and "clean" the point cloud as much as I can);
  2. When I'm satisfied with the result I import the vector data that needs a Z value and use the "Apply elevation to the selected features" (this might not be correctly translated because I use GM in my native language).

As I said, everything worked fine on previous projects but this one is giving me headaches. On this project I get a lot of "bumpy" roads and peaks, like the image bellow:

In some cases, I've noticed that it's related to the 2D drawing (my co-workers draw in CAD on top of orthophotos) that is a little bit of the road so I get peaks from points like trees or high noise that I didn't clean, but those situations are normally a small volume and easily solved. However, the volume of data I work on doesn't allow me to verify manually and individually every case so I try to automate as much as I can and focus on the more notorious problems. In the case of this project I get this peaks and bumps all over the project.

The only solution I'm thinking of is removing vertices in CAD so I can make the roads more homogeneous without changing the real elevation of those vertices. However, this would be very time consuming and I don't feel like I'll be able to get to the result I want.

Have any of you had problems with this or have any ideas that I might try to solve this?


Obs: This is a Riegl Las, and they have been, by far, the ones I get the best results with. This was also classified by A.I.. I've been working like this for a while and it usually turns out very good.


Thanks in advance!

Best Answer

  • DerrickW
    Answer ✓

    Try binning your lidar data, after going through noise removal. Depending on the site, MIN or MEDIAN might work best (bin settings in the elevation grid creation options, not lidar thinning).

    Tops & Toes is where the binning process is going to affect your data most/first, so take a close look at those spots as you're iterating through different bin settings.

    If your ground-classified point cloud is relatively free of noise and residual vegetation, I'd start with something like Binning MEDIAN with a 1ft spacing. Once you get up to about 3ft-5ft is where you'll start seeing it affect TOPs, in my experience.

Answers

  • Hello Derrick,


    I'm sorry for the late feedback. Unfortunately, by the time you answered this topic, I already did tried what you suggested, with no luck though. The project has been completed but it required a lot of manual work.


    For those interested in the solution, there were no rules, I just messed around with both surfaces generated by GM and C3D. After trial and error, I ended up using a very bad surface, with very bad roads and, after smoothing the roads manually, I adapted the surface to those new roads and then did all the work on the surface itself. It ended up working out fine, for the time I had to complete it. However, I'm looking forward to work a little bit more on it in the future.

  • Just came across this post as I have a similar problem. In my case, I have terrestrial scan data which has not been properly registered. This was partly caused by the scan stations being too far apart for the density of the scans. It seems that the software is able to do a decent job aligning the scans horizontally, but not vertically. We have tried to fill in some longer gaps with more scans, but registering the scans together and getting good results continues to elude us. I will be working with our supplier to see if they can help salvage the data we have. Otherwise we may just have to re-scan the project.

    GM can smooth data, but if the source data is inaccurate, the result will simply be inaccurate data that looks better than it really is. It may be good enough for the project requirements, but not what it could be.

    As for equipment, I have used data from Leica P series scanners as well as Riegl and a Trimble. All can give great results, but I have had occasional registration issues with all of them.