I need to obtain a grid of bare ground elevation (removing all edifices, trees, roads and man-made elements) for quantitative geomorphological analysis from a high resolution (1 m) DSM obtained by orthoimages (not from LIDAR). Is there any suggestion you can provide me to filter my original grid?

  • do you have any information on the vertical features ?
    – radouxju
    Jun 17, 2014 at 14:53
  • I have this 1-m resolution DSM and the corresponding orthophotos, tile 2,5 km x 2,5 km each. I can clearly visually distinguish buildings, road and tree tops from geological-geomorphological features but I do not know how to automatically remove those features extruding from ground surface
    – geoclaudia
    Jun 17, 2014 at 15:02

1 Answer 1


This filtering process is usually performed on a lidar point cloud and not an interpolated derivative. It is unlikely that you will have satisfactory results attempting to filter the DSM. I would highly recommend tracking down the original lidar data.

You could attempt to treat your DSM as a point cloud by converting it to points and then running a filter intended for lidar point clouds. Depending on the algorithm, you may get a suitable result for generating a bare earth DEM. However, it may end up oversmoothed and not supporting the current resolution of your DSM.

Some recommended "free" lidar filtering software:

Airborne LIDAR Data Processing and Analysis Tools (ALDPAT)

GRASS GIS specifically, v.lidar.correction

Idaho State University Boise Center Aerospace Lab IDL Virtual Machine software (BCAL)


USFS-PNW lidar processing and visualization software (FUSION)

USFS-RMRS Multiscale Curvature Classification (MCC)

  • 1
    Would it be possible to filter out "items" that have certain features - like, up to this size wide and up to this size tall, but not shaped like this (for instance, as a cone?)"? I know I am asking for something quite complicated...
    – Dakatine
    Jun 17, 2014 at 17:53
  • 2
    "DSM obtained by orthoimages not from LIDAR"
    – Martin F
    Jun 17, 2014 at 19:59
  • Hi all, thanks for you replies, that will be really useful to me IF I can obtain LIDAR data of my area. What I have now is not a LIDAR, but a cool othoimage-derived 1-m resolution DSM. The only problem is that it is not what I needed for my quantitative analysis pixel by pixel :( I was thinking about a tedious work of removing "by hand" the unwanted object, i.e. delimiting with polygons areas with trees, areas with buildings etc, assigining them an average thickness, converting to grid and subtracting from my DSM, but the result would be poorer than a countour-derived 10-m DEM I already have.
    – geoclaudia
    Jun 18, 2014 at 14:43
  • 1
    My point was that you could try applying filtering methodology developed for lidar point clouds to your DSM after converting them to points. You would in essence, have a systematically sampled point cloud and some filtering algorithms may work. I believe that my MCC algorithm would work if you played with the model coefficients. Jun 18, 2014 at 17:13
  • Dear Jeffrey, thanks again. I'm reading your 2007 paper, and will try to do what you suggested. This will be useful also if I obtain LIDAR data. One question: MCC algorithm will work also on edifices or only on forested areas?
    – geoclaudia
    Jun 19, 2014 at 10:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.