[Openstreetmap-dev] Re: [Openstreetmap] Landsat update

Nick Whitelegg nick at hogweed.org
Sat Apr 30 18:33:39 BST 2005


Jo,

> re nick,
>
> On Sat, Apr 30, 2005 at 01:03:00PM +0100, Nick Whitelegg wrote:
> > I'm wondering if it's to do with the inaccuracy of WGS-84 at the UK's
> > latitude
>
> no, apart from the usual GPS error factor it shouldn't make a difference.
>
> > i.e. it uses Airy coordinates even though the GPS uses WGS84.
> > to screen display, the error becomes worse!
>
> http://www.free-map.org.uk/drawmap.php?e=490806&n=130314&mapview=10
> what's being passed in here is OSGB projected numbers, yes?
> i now feel quite confused as to what you're reprojecting where :)

In freemap, the track and point coordinates are stored internally using OSGB 
gridrefs (yes, maybe not the best design decision but anyway...) and 
displayed without need for complex transformation. They were originally 
surveyed with a GPS (WGS84).

The SRTM data are stored as lat/lon and transformed to screen points using the 
same transformation that I used to transform the GPS points.

>
> agreement with this. quick consultation with Schuyler, couple of
> suggestions as to what the problem is:
>
> - with SRTM at 90M and Landsat at 15 or 30, it's possible that the
>   appearance of accuracy over the SRTM layer is a lucky coincidence?
>   or do your other experiences bely that?

Haven't had any inaccuracies with the SRTM which are as noticeable as the 
Landsat... at a guess I would say the Landsat is perhaps a little under a 
quarter of a mile (perhaps 300m) out.

> - the projection error could well be on the landsat WMS side;
>   "misregistered" or slightly misrectified tiles. Can you try this out
>   against a different set of known good data and check the fit?

Haven't got any other sources of aerial data, but have tried three areas (west 
Sussex, Southampton Water and the New Forest) and consistently get about the 
same error.

Not really used irc before I have to admit.... but am quite happy to keep 
talking via email.

Nick




More information about the dev mailing list