[OSM-talk] The long tail - lowest common denominator

[Lars Aronsson]

Simon Hewison wrote:

> You have just done a perfect demonstration about just how coarse 
> the VMAP0 dataset is, which I think is due to it being designed 
> for rendering maps at 1:1000000, rather than the sort of scales 
> we tend to concentrate on in openstreetmap.
> 
> For instance, it has some lakes and reservoirs, but some are 
> over 200 metres away from where they actually are.

Perhaps one could use the coarse vector hydrography of VMAP0 as a 
guide for interpreting nearby dark features in the Landsat images? 
One could run contour detection (between dark blue water and green 
or brown fields) on the Landsat and vectorize this contour, but 
this vector data would also contain lots of irrelevant lines.  
Then one would only keep only those contours which are within, 
say, 400 metres of the VMAP0 hydrography.

I think the error can be larger than 200 metres at detailed river 
bends, such as this one, 
http://london.freemap.in/osm.html?lat=8019111.94824&lon=1739910.91992&zoom=13


-- 
  Lars Aronsson (lars at aronsson.se)
  Aronsson Datateknik - http://aronsson.se