Global Mapper v25.0

Classifying LiDAR data on a landfill

jmonell
jmonell
in Elevation Data
I captured 2 square miles of data with a fixed wing.  Pointcloud was generated in Pix4D, now looking for the best way to classify the ground data for DTM generation.  Very dense point loud, metadata says between 90 and 120 points per square meter.  I've tried several different parameters on a subset, guessing I need to run the classification tool, check, turn off the ground points, adjust params, and run again.  Typical features include dense woods, low shrubs, concrete barricades, vehicles, and gravel stockpiles.  Any help/advice would be appreciated, thanks.
Tagged:

Answers

  • mudfoot
    mudfoot Global Mapper User
    You do have a dense point cloud so you may first try 3 Point Spacings as your Base Bin Size, which may speed up processing time.

    The attached file shows the settings I recently used for some rolling to very steep terrain.  The point cloud density was approximately 9 points per square meter and the terrain contained a good amount of low vegetation, which is why I specified the 'Minimum Height Departure" of 0.2m in an attempt to get through the low vegetation.  From my limited experience there appears to be a fine line between identifying non-ground points, like low vegetation, and losing definition of hard terrain breaks such as ridge lines.

    After running auto classify to ground I plan to use a mapping vector file of shapes to identify structures such as bridges, buildings, etc. and use the shapes to select all the LiDAR points within and reclassify them as buildings.

    This is all new to me too so take from it what you can or toss it.  There isn't a great deal of help regarding this so just test smaller sections to see what works for you. Good luck. 


  • jmonell
    jmonell
    Answer ✓
    Thanks for the reply.  GM's doing much better than my UAV software, but finding the sweet spot is a pain.  Agreed, help is spotty at best.  I tried variations of your recommended settings, I think the best bet is to run autoclass ground, turn off the points it finds, adjust, and rerun, repeating until ground points are classified.  Again, much more efficient than Pix4D.

Categories

In this Discussion