RSME comparing LIDAR data with a third party’s 3D dataset

I would like to share with you an easy analysis i have been working in the last days. I had a vector dataset of buildings and i knew how high they were (there was a field called ‘AGL’ or Above Ground Level) and a LIDAR 2m resolution dataset over the city of London. My aim was comparing both sources, understanding LIDAR data was the actual reality (or a closer version to it) and my source of 3D buildings was the dataset i needed to deliver to my customer… Te actual height of those 3D buildings had been extracted using stereo photogrammetry methods. I also needed to focus on residential data, so heights below 15m… So make it easy. The question was:

How accurate is my dataset of residential buildings over London?. Which is the RMSE measuring them both?

I used Global Mapper v.13.2 (b062012) and ArcGIS 10.0 (b3200)

This is the 2m resolution LIDAR data provided by geomatics-group.co.uk

I also needed to get a layer of points out of this dataset so i used Global Mapper and went to Files/Export elevation grid format and choose ASCII as the format.

This is the layer of buildings and their AGL as label

I flagged those residential buildings

and using ArcGIS i performed a Spatial Analysis using Arctoolbox/Spatial analysis to join the Lidar heights in ASCII format and the residential heights… to be able to measure the difference between both datasets

this way i got a new vector layer which table contained both elevation fields (Lidar and my 3D buildings)

As you can see, i added a new field in ArcGis using table/add field and added ‘compare’ and SQL [“AGL”- “ELEVATION”]

then i measured it visually using a density grid in Global Mapper. Create density Grid.

And finally measured the RMSE by opening the table in excell format and usign the actual formula for extracting RSME values:

= SQRT(SUMSQ(M1:Mn)/COUNTA(M1:Mn)) —> Note this formula is only valid for this case. You’d need to update Mx values using yours:-)

Wow! a very high value. Does this value corresponds to our accuracy figures? Yes? No?.

Now it’s the time for decission makers to bring into action!

LIDAR-08

And what about some geostatistical analysis. I performed this using North East Trends in ArcGis. We can see from West to East there’s no variation but we can see it increases the error the further the south…