Solved

Resampling a raster using the minimum value

  • 26 April 2021
  • 5 replies
  • 11 views
olivier
Rank
Badge +1

Hi,

 

I make an attempt at removing the effects of building heights on a DSM (Digital Surface Model) in order to obtain a result looking more like a DTM (Digital Terrain Model), in other words something closer to the bare soil elevation. My thought was to use a RasterResampler transformer and define a cell size a bit larger than the buildings, and also assigning the minimum elevation value to the new pixel. However I don't see a minimum value method for RasterResampler. I looked at the RasterConvolver but there is no way to define the kernel size in terms of distance instead of pixels. Correct me if I'm wrong. In other words I need to find the minimum over a great number of neighboring pixels (much more than the traditional kernel sizes).

How would you achieve such a raster operation using FME Desktop?

 

Thanks in advance!

 

Best regards,

Olivier

icon

Best answer by caracadrian 27 April 2021, 06:34

View original

5 replies

caracadrian
Rank
Badge +20

I recommend transforming your raster to pointcloud via PointCloudCombiner, classify the ground via LAStools.lasground (from the FMEHub), then recreate your raster from the filtered data if you need to.

virtualcitymatt
Rank
Userlevel 4
Badge +26

I recommend transforming your raster to pointcloud via PointCloudCombiner, classify the ground via LAStools.lasground (from the FMEHub), then recreate your raster from the filtered data if you need to.

Just keep in mind that lastools should be licensed unless for non-profit or educational purposes.

olivier
Rank
Badge +1

Thank you @caracadrian​ !

Following your recommendation, I am now evaluating LAStools solution. I am trying to figure out why I am getting errors in the log when running the LAStools.lasground in the FME workspace:

ERROR: cannot alloc tile

ERROR: cannot get tile (3000/4363)

ERROR: cannot alloc LASrasterTile

 

I have of course transformed the GeoTiff raster DEM into a point cloud using the PointCloudCombiner and I also reprojected it to a coordinate system in meters before submitting it to the LAStools.lasground (pointing to the correct exe).

As LAStools is a tool which is not free, can you think of any alternative method?

 

caracadrian
Rank
Badge +20

LAStools is the simplest and most memory efficient way of processing this particular issue via FME.

To workaround your errors you should play with the transformer parameters (IE tile size, number of points).

An FME only solution would be using RasterSlopeCalculator, building buildings footprints polygons from a range of pixel slope values (maybe add a palette), clipping the raster by the buildings footprints (maybe add a buffer to them), creating a new DEM from the clipped raster with the holes filled with the minimum value of the surrounding area (the buffers come in handy here).

The memory usage of this workflow is greater then the LAStools one so use precaution on processing large DSM's.

olivier
Rank
Badge +1

LAStools is the simplest and most memory efficient way of processing this particular issue via FME.

To workaround your errors you should play with the transformer parameters (IE tile size, number of points).

An FME only solution would be using RasterSlopeCalculator, building buildings footprints polygons from a range of pixel slope values (maybe add a palette), clipping the raster by the buildings footprints (maybe add a buffer to them), creating a new DEM from the clipped raster with the holes filled with the minimum value of the surrounding area (the buffers come in handy here).

The memory usage of this workflow is greater then the LAStools one so use precaution on processing large DSM's.

Yes I will play some more with LAStools parameters. I am new to 3D data but the tool seems promising when looking at what it leads to when using sample data. If it meets our requirements we will contact rapidlasso for buying a license. I also read it adds some noise to the evaluation results, so maybe results with the license will look even better.

The FME-only solution is a good B-plan. The trickiest part though is the extraction of building footprints. But I think we can do some magic there. Furthermore we have polygons of built-up areas, so removing these larger parts from the DEM is an option (quick and dirty), followed by refilling the holes with the neighboring values.

Reply


Community Stats

28,669
Posts
109,357
Replies
36,531
Members
Terms & ConditionsCookie settings
A product of
Safe Software respectfully acknowledges that we live, learn and work on the traditional and unceded territories of the Kwantlen, Katzie, and Semiahmoo First Nations.