Skip to main content
Question

Extract Building footprints from GEOTIFF Raster images

  • February 28, 2022
  • 3 replies
  • 439 views

I'm trying to extract building footprints from GEOTIFF raster images. I have raster Orthophotos (Aerial imagery), Digital Terrain Model (DTM/DEM) and a Digital Surface model (DSM), but I do not have the building footprints.

 

My end goal is to use the DEM + DSM to extract the height of the buildings and extrude the building footprint to 3D models.

 

I found this article: https://smartcarto.com/gis/detect-buildings-in-satellite-imagery-using-machine-learning-fme-vs-arcgis-pro/, but it doesn't really go into detail, and since I also have DEM and DSM, not just orthophotos I thought I migh thave other possibilties. I've extracted building footprints from Point Cloud date before, but this time I only have acess to Rasters.

 

Is this possible in FME? Is there any tutorials?

 

Thanks in advance.

3 replies

hkingsbury
Celebrity
Forum|alt.badge.img+63
  • Celebrity
  • 1625 replies
  • February 28, 2022

If you have the DSM you could potentially pull out the rough building footprints by looking at where elevation values jump a significant amount. This would indicate a vertical surface (side of a building) it would be pretty dependent on the resolution of your data, i'd be wanting sub meter data to undertake something like.

 

Alternately have you had a look at Open Street Map data for your area? They have pretty good coverage.

Also have a search through GitHub for Building Footprint extration, there are quite a number of public projects that people have shared that provide tools to achieve that task


  • Author
  • 4 replies
  • February 28, 2022

Thanks for the reply. I've already looked at OSM and other sources with no luck. I'm working on a smaller city in Albania, and the gourvernments official website does not have footprints for this area either.

How would I go about locating elevation jumps in the dsm and getting shapes from that? I haven't worked with geotiffs like this before, but I guess I would have to extract elevation values for each pixel somehow? The resolution is 0,5 meters, which should be enough to get some results. The orthophotos are even higher resolution if I could somehow use those, or a combination. I know ArcGis has some tools for this, but I don't have a license.


maayke
Contributor
Forum|alt.badge.img+6
  • Contributor
  • 18 replies
  • April 22, 2022

Hi @robinosc​, I'd use a RasterExpressionEvaluator to compare the DSM and DTM. Following this comparison, positive cells should indicate above-surface objects. This should leave you with a raster datasets with positive cells for for example trees, buildings and son. After this, there are multiple ways to continue: you could use e.g. the orthophoto, area calculations and so on.