I've got a copy of some old 25inch to the mile maps for my area and I've
compared the new StreetView maps with this to see if it's worth tracing
buildings. Almost all the detail of actual building shape is lost in
StreetView with most buildings reduced to a simple rectangle. Some are also
half "missing" as they are covered up with a road which is obviously drawn
on after the buildings are. However it does seem useful for locating the
positions of buildings. I wouldn't have thought it was worth bothering in
most cases.

Another potential source of building shapes is photogrammetry, this is
something I've been trying in the middle of Market Harborough with St
Sionysius which produced some good data. If you want to see exactly what I
did for that take a look at my OSM diary. Incidentally, St Dionysius in
streetview is just a rectangle. For public buildings with good access all
around this is definately a viable data source (if somewhat time intensive
at the moment but we're working on this).

Ainsworth

On 5 April 2010 17:57, TimSC <mapp...@sheerman-chase.org.uk> wrote:

>
> Hi all,
>
> I was looking at the available OS data and one of the things they have
> but is not really in the current OSM database is building shapes.
> Individual buildings appear in the Street View data. The VectorMap
> District data seems to focus on land use and only has "important"
> buildings. The Street View data is much more complete in terms of
> individual buildings. I guess the questions we have to address are: "is
> this data worth having in OSM?" and "are there any better sources of
> building outline?" I think it would be a valuable addition for highly
> detailed mapping of urban areas, campuses, etc.
>
> The other sources of building outlines I know of are the 1st edition OS
> sheets (which are rather old), Yahoo imagery (which takes work to
> trace), and manual surveying (which is a major hassle since buildings
> tend to throw off GPS positions). Another source would be preferable to
> these.
>
> If we do want this data, it seems to be only available as a raster
> layer. (Does anyone know if this will be available as a vector layer?)
> Converting raster map data to vector data is non trivial but I think it
> is achievable. The building colour in the raster map seems to be unique.
> The raster maps could be selected by colour, vectorised and simplified
> (to remove redundant nodes). Since we have a significant number of
> buildings mapped in the UK, we could presume that we don't want to
> overwrite or duplicate what is already there. Buildings could be
> imported if it would not overlap with an existing building in OSM. One
> problem is some buildings are marked as leisure=stadium without a
> building=*. We might end up with duplicates of stadiums, leisure
> centres, churches and so on. Would we want the import to be more
> cautious to prevent this? I won't be importing anything until the
> rectification of Street View is complete (and I get a feel for the
> consensus on the mailing list) but I have demoed most of the method in
> python using the shapely module to do some geometry transforms.
>
> Any general thoughts on buildings and imports? Will this data be
> officially available in vector form at some point?
>
> We don't seem to have much comment on the legal issues and the
> implications of ODbL yet. Do we have a view as to if our attribution is
> sufficient?
>
> Regards,
>
> TimSC
>
>
> _______________________________________________
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>
_______________________________________________
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Reply via email to