Solved

TIN Surface to Raster - Quality problems with RasterDEMGenerator

  • 17 January 2022
  • 2 replies
  • 23 views

Badge +9

Hi,

 

I am having problems with converting a TIN surface to a raster with height information. The input data is a LandXML which contains a surface of type IFMECompositeSurface. Right now I feed the surface into the RasterDEM generator with the following settings:

RasterDEMGenerator SettingsThe coordinate system is EPSG:28992 so the surface tolerance is set to 1cm, however the results are quite jagged:

RasterDEMGenerator Result 2DTo give you a sense of scale, the image above is for an area of around 1300 meters wide so the the jagged points seen above are sometimes a 100 meters wide.

 

To check my results I decided to convert the result to a pointcloud so I could take a look and compare it with the original surface in the 3D view in FME Inspector. In the image below you can see that the pointcloud that was created from the raster actually dives beneath the original surface:

Pointcloud dives under original dataDoes anybody have an idea what could cause these problems?

I a doubting whether the RasterDEMGenerator is the right transformer for this job. I already have a workaround where I use the PointCloudCombiner and a NumericRasterizer for a much nicer and smoother result but that method has a problem dealing with very small faces which in the end cause holes to appear in the raster.

icon

Best answer by daveatsafe 18 January 2022, 00:43

View original

2 replies

Userlevel 2
Badge +17

Hi @birgit​,

The RasterDEMGenerator generates a new TIN, treating the input surface as 3D points for the triangulation. This can create artifacts at the edge of the DEM, where concavities are bridged by new triangles.

I think your method of using the PointCloudCombiner, then NumericRasterizer is the best method. Perhaps try again with a smaller point interval for Vector Feature Interpolation.

Badge +9

Hi @birgit​,

The RasterDEMGenerator generates a new TIN, treating the input surface as 3D points for the triangulation. This can create artifacts at the edge of the DEM, where concavities are bridged by new triangles.

I think your method of using the PointCloudCombiner, then NumericRasterizer is the best method. Perhaps try again with a smaller point interval for Vector Feature Interpolation.

Hi @daveatsafe​ ,

Thanks for the explanation on how the RasterDEMGenerator works. I already had a suspicion that I wasn't using the transformer in the correct way and your answer has confirmed that for me. I have upped the interval for the Vector Feature Interpolation and the results from the NumericRasterizer are pretty nice. Unfortunately the workbench is slowed down with the 4 times increase in points so I am going to check if parallel processing is going to help in this case(or just make it slower).

Reply