[OSM-talk] Suggestion more complete mapping verifactiion

[Dair Grant]

John Baker wrote:

>You give me the list of "official" street names in a given area and I will
>do the rest.

Aside from the licence issues, the error rates for "official" 
street lists are surprisingly high.

I have almost finished mapping Haywards Heath 
(http://tinyurl.com/2dxh5p), which has a population of 20K and 
covers about 10km^2. Google's map has plenty of errors on it, 
and they buy this data from TeleAtlas (who probably have the 
best commercial road network for the UK).

You can find streets with the wrong name, which are in the wrong 
place, which connect when they shouldn't, are on the map when 
they don't exist, exist but aren't on the map, are mapped as a 
road but are actually a private driveway, etc.


I don't think having a "definitive" list of street names for an 
area is going to be as useful as you think, as short of going 
out in person you have no way of knowing how accurate that list is.

I have a geo-tagged photo for every street sign in Haywards 
Heath, but if I sent you those then how would you use that information?

The best you could do is estimate that a lot of street names in 
an empty area of the map probably mean that area is 
under-mapped, but that doesn't actually improve the map.


I have no idea how useful it would be in practice, but a better 
way (no licence issues, works outside the UK) to find those 
sparse areas might be to use an image-based technique.

Take some tiles from OSM and another service, recolour then so 
you just have road/background colours, blur a couple of times 
then subtract from each other, and look for large areas where 
there's "stuff" in one map but nothing on the other.

My guess is it won't tell you anything your eye couldn't tell 
you any way, but it might help you identify holes in otherwise 
"complete" areas.


-dair
___________________________________________________
dair at refnum.com    http://www.deathvalleycycle.com/