[OSM-talk] Fwd: The completeness of OpenStreetMap

[Richard Fairhurst]

Edward Johnson wrote:

> This looks like quite a good system and at least it breaks it down   
> into which categories are complete and which not. This system though  
>  relies on users verifying whether an area is complete. I would   
> really like it if we could have a model/algorithm that could tell us  
>  which areas are compete and which not. Any ideas?

First step (I think) is to analyse the _likely_ density of  
roads/buildings/whatever in a given km square - then compare OSM  
coverage within that square against that.

Things you could use to find out the density:

- Colour analysis on aerial imagery (e.g. Landsat)
- Terrain - watery/hilly areas less likely to be dense
- Compare against other free datasets (e.g., for the UK, density of  
postcodes on NPEmap)

cheers
Richard