[Imports] Addition of building footprints in selected U.S. and Canadian cities

William Morris wboykinm at geosprocket.com
Thu Mar 22 20:03:47 UTC 2012


Hi Ian,

Here's a rundown of the methodology: http://bit.ly/GRgAqO (the same
was used for the rest of the sites on the coverage map -
http://g.co/maps/h9qfx).

Object-oriented classification is very powerful, but also somewhat
labor-intensive and computing-resource-heavy, particularly at this
resolution. And in each case, the building detection is occurring
through a combination of multispectral image segmentation and
LIDAR-based topographic extraction. While the Bing imagery available
to us in the editors is at a great spatial resolution, it's not
sufficiently detailed to capture building edges using the same methods
as the layers here. That's not to say I don't think it's possible to
create an editor plugin that runs a generic segmentation algorithm. A
future project . . .

On the other notes, the files are currently in .img format, but it's a
snap for me to convert them to vector .shp and then run shp2osm or
something similar. I don't have a coverage showing complete extent of
the sites, but in most cases they are limited to the municipal
boundary (i.e. the city of Des Moines). I've looked at some of the
sites and seen some existing buildings - particularly in urban
downtowns - but this would provide comprehensive building footprints
for these areas. I would run a complete conflict analysis before any
import.

-B


----------
William Morris
Cartographer
(802)-870-0880
wboykinm at geosprocket.com
Twitter: @vtcraghead

GeoSprocket LLC, Burlington VT
www.geosprocket.com



On Thu, Mar 22, 2012 at 3:47 PM, Ian Dees <ian.dees at gmail.com> wrote:
>
>
> On Thu, Mar 22, 2012 at 2:39 PM, William Morris <wboykinm at geosprocket.com>
> wrote:
>>
>> Hello all,
>>
>> Some colleagues at the UVM Spatial Analysis Lab have offered to submit
>> features to OSM that are derived from their high-res (1 meter) LULC
>> analyses at locations throughout the U.S. and around Toronto
>> (locations here: http://g.co/maps/h9qfx). Their work was primarily
>> intended to map urban tree canopy, but they've captured various
>> impervious surfaces as well, including excellent building footprints.
>>
>> Here's an example near UMD of current OSM coverage vs. UVM-SAL
>> coverage: http://flic.kr/p/bseokW
>>
>> Unless there is a tremendous need for tree canopy coverage in OSM, I
>> think it would be most useful to add just buildings, and possibly
>> parking, both of which are separable feature-by-feature. I'd be happy
>> to screen the imports for conflicts with existing OSM features, and to
>> bias toward those existing features in all cases.
>>
>> Given that this is an import of several thousand features, what is the
>> best way to proceed?
>
>
> Building footprints are interesting and we an talk about importing them a
> bit more, but can you describe how they were derived? It looks like there
> was some classification algorithm used to spot tree cover, parking, roads,
> and buildings. Is the building detection fast
> enough/compartmentalized-enough to be thrown into an editor plugin so we
> could do that automatically from within our editors?
>
> As far as importing, can you describe what format the files are in
> currently? Exactly which areas do they cover? Have you looked at those areas
> to see if there is existing data?



More information about the Imports mailing list