Pit and Spike removal from DEM

KBlocker

Is there a process for individually removing pits and spikes from DEM that will not affect the surrounding terrain elevations?

Thank you

global_mapper

What you could do is right-click on the DEM layer in the Control Center and select to create points at the grid cell centers, then edit the elevations of those points in pits or spikes (or just delete those points), then right-click on the point layer in the Control Center and select to create an elevation grid from that. The pits and spikes should then be removed or modified based on your changes to the points.

Let me know if I can be of further assistance.

Thanks,

Mike
Global Mapper Guru
gmsupport@bluemarblegeo.com
http://www.globalmapper.com

KBlocker

I have a rather large area to work on, is there a way to create the control points just in the area of the pits/spikes?
forgive me if this seems like a rtfm moment.

global_mapper

There isn't really a way to automatically identify the pits and spikes. I suppose you could load the layer twice, then set one to use a resampling method like a box averager of an appropriate size, then use File->Combine Terrain to create a new difference layer and easily find those places where the terrain deviates significantly from the local average. Nothing really fool-proof though.

Thanks,

Mike
Global Mapper Guru
gmsupport@bluemarblegeo.com
http://www.globalmapper.com

RonTheSurveyor

Just stumbled on this when looking for the same thing.

After thinking about it for a while, I figured out and tested a process.

I created a surface from noisy lidar data which had many spikes. I then created contours at 0.25 m interval and intentionally kept small closed contours. I converted those to areas and then selected all ground-classed lidar points that were inside an area less than 0.2 sq metre. It took a while, but about 40,000 lidar points were detected which I classed as low vegetation. I then created a new surface with the resulting ground classed lidar points using the same settings as before. Results are shown below. Processing time was significant as original point cloud has 20M ground classed points. I would have achieved better results if I used a smaller contour interval (perhaps even 5 cm) and or filtered by larger areas as many spikes remain on the road surface. But this was just a proof of concept.