[OSM-talk] How can we measure the completeness of the map statistically

Freek freek_osm at vanwal.nl
Fri Jun 27 17:52:28 BST 2008


On Friday 27 June 2008, Edward Johnson wrote:
> I'm very intrigued by your work Freek. 

Thanks :-)

> I'll get working on trying to reproduce something like this for the UK

If you can wait for a few days, I can make some for the UK too (busy finishing 
my exams a.t.m.).

> and then try to use the map 
> generated to compare with OSM. See if the hypothesis does hold that more
> complete areas are edited more. Because of course as you say a fully
> complete area should not be edited at all.

Difficult... In the Netherlands it might be easier because we have the AND 
data as a "background" to compare to: the number of edits in different areas 
can probably be compared by dividing them by the amount of AND data for the 
areas, which is again approximately what I'm doing for the blue-green-red 
maps. For other areas you really need another external source to compare to. 
Some ideas:
- VMAP0 has some city boundaries I remember (and cities also have a special 
colour on landsat imagery); data density in cities will be higher than in 
rural areas, on average.
- Some governments publish road-length statistics, someone used these to 
generate a top-10 of most complete municipalities before the AND import:
http://wiki.openstreetmap.org/index.php/WikiProject_Netherlands/Statistieken
Interestingly, these statistics generated some healthy competition among Dutch 
participants at the time.

I think we should just work out some of the different methods (from fully 
automated based on different sources to manual tagging) and see which work 
best. Probably manual and automated methods will just coexist in the future.

-- 
Freek




More information about the talk mailing list