FW: Landsat ? (was:Re: [Openstreetmap] New roads appearing withoutGPS tracks)

Etienne Cherdlu openstreetmap-L at gj0.net
Fri Dec 16 14:44:59 GMT 2005

As far as I can tell from looking at GPS tacks and Landsat images of the UK,
the Landsat images are not as accurate as the GPS tracks, but are still
quite good for purpose.

Compare a Landsat image of a Motorway with a single GPS track for the same
motorway.  The Landsat might be 20-30m off, but then you may not know which
lane or side of the motorway was being driven when the GPS tracks were
laid.  Given that motorway can be easily 30m wide its hard to judge that one
may be more accurate than the other.

Add to this the fact that the motorway path may be curved.  If not enough
nodes are drawn then the error introduced due to the segment being a
straight line can also easily be 20m or more.

If the Landsat is clear enough to pick out the route of a road then I don't
think any harm is done by mapping it even in the absence of any GPS tracks.
If GPS tracks are later added the path can be reviewed and refined in the
true collaborative Wiki manner.

By comparison, does anyone know how accurate the roads on Google maps are?


*From:* Mikel Maron [mailto:mikel_maron at yahoo.com]
> *Sent:* 16 December 2005 12:25
> *To:* openstreetmap at vr.ucl.ac.uk
> *Subject:* Re: Landsat ? (was:Re: [Openstreetmap] New roads appearing
> withoutGPS tracks)
> From: Tom Carden <tom at tom-carden.co.uk>
> The vast majority of OSM data so far has been entered with the
> assistance/sanity check that landsat offers us.  We'd have to junk the
> lot if there was ever a successful challenge on derived landsat data,
> but I don't think there ever will be.
>  <http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap>Landsat 7
> is in the public domain. There is no restriction on derived works.
> However I would be cautious, when relying on Landsat for drawing vectors.
> Only usable as a very rough sanity check (ie, generally in the right part of
> the world). Until there are other imagery sources integrated, GPS traces are
> the primary vehicle for deriving vectors. Don't think there's any
> disagreement on this, just want to make the point.
> There are serious rectification errors in the JPL Landsat data. For
> instance, the shoreline here in Brighton is off by 20-30m.
> Some discussion on this in the following WorldWind thread (post #8)
>   http://forum.worldwind.arc.nasa.gov/index.php?showtopic=1512
> -Mikel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20051216/55ee6158/attachment.html>

More information about the talk mailing list