Digital Surface Model

Question

Hi all,

Probably this discussion has been risen before. I have a Digital Surface Model (DSM) but I need to "clean" it as I am not interesting neither on trees nor bushes. I need it as the DSM will represent the height and position of the buildings or bridges more accurately. My question is about the process about how to "clean" or filter the data in order to have the information that I require.

Kind regards,

Miquel Vinyals

XanderBakker · Answer

Hi Miquel Vinyals‌,

My first advise would be to go back to the source (if possible). If the source is LIDAR (and it looks like it) the LIDAR files (las files?) have information on the return. This is normally used to derive the DEM (of DTM).

You may want to have a look at a presentation I did on a project that is simular. It may contain some pointers to do this: Detecting Buildings in AHN2 LiDAR data with ArcGIS - Grontmij

Kind regards, Xander

MiquelVinyals · Answer

Hi Xander,

Thanks for your quick reply, please, find a picture of a DSM and DTM. As you can see, on the DSM we can find lots of trees and bushes that I am not really interested on them, just the buildings. In theory, that information is coming from a LiDAR.

The Data is on a Grid Format (Cellsize 1,1 being continuous source type).

Hope it helps for your advise

Looking hearing from you,

Regards,

Miquel

XanderBakker · Answer

First of all, there may be some discussion on the definition of DSM, DEM and DTM. There is a nice discussion here: What is the difference between DEM, DSM and DTM? - Geographic Information Systems Stack Exchange and I have a presentation here: Digital Elevation Models - WUR - Grontmij

Before we can suggest a method to filter or clean the data to exclude vegetation, it is necessary to know what the source of the data is (how was it generated, what resolution does it have, how large is your extent, etc). This may influence the procedure and/or in some cases (if the source is LIDAR) lead to the suggestion to go back to the source and create a raster based on the last return...

If possible include a small example of the data or a picture of what you have.