GPS track oddities

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

GPS track oddities

Jonathan McDowell
I've been trying to produce some tracks around Norwich, but have been
having problems with my tracks appearing out compared to the Landsat
maps. At first I thought this was due to my cheap USB GPS receiver, but
I tried with a Magellan hand held device as well with similar problems.

So I tried something else. I used Streetmap to look up co-ordinates of a
roundabout (52.645225 1.276237) and a bit further up that road
(52.645992 1.273559).

Looking at the points with Google Maps shows them as you'd expect:

http://maps.google.co.uk/maps?q=52.645225+1.276237&iwloc=A&hl=en
http://maps.google.co.uk/maps?q=52.645992+1.273559&iwloc=A&hl=en

So I created a GPX file with these 2 points in it:

<gpx version="1.0" creator="A quick Perl hack"
        xmlns="http://www.topografix.com/GPX/1/0">
<trk>
<trkseg>
<trkpt lat="52.645225" lon="1.276237">
<time>2005-10-14T17:40:02Z</time>
</trkpt>
<trkpt lat="52.645992" lon="1.273559">
<time>2005-10-14T17:41:02Z</time>
</trkpt>
</trkseg>
</trk>
</gpx>

and fed this into osm-editor. Tell it to pull the Landsat data and I get
an image like:

http://the.earth.li/~noodles/osm-wrong.png

(248k)

This seems to me that I'm not going insane and that there's something
wrong with either the LandSat data or the way osm-editor and OSM deal
with it. Can anyone shed any light on this?

J.

--
    Love is an attraction to a     |  .''`.  Debian GNU/Linux Developer
  perfectly normal person you've   | : :' :  Happy to accept PGP signed
  temporarily mistaken for a god.  | `. `'   or encrypted mail - RSA +
                                   |   `-    DSA keys on the keyservers.

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

Re: GPS track oddities

Nick Whitelegg-2
Hello Jonathan,

Are you zoomed in too far? Landsat is only low resolution data so if you
zoom in too far it becomes very indistinct. It's better for viewing say 5km
x 5km areas. The main purpose for incorporating Landsat into osm-editor is
for people to estimate polygons such as lakes etc, though it also provides
interesting browsing over large areas.

Nick


_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

Re: GPS track oddities

Lars Aronsson
In reply to this post by Jonathan McDowell
Jonathan McDowell wrote:

> I've been trying to produce some tracks around Norwich, but have been
> having problems with my tracks appearing out compared to the Landsat
> maps. At first I thought this was due to my cheap USB GPS receiver, but

The Landsat images are even cheaper than your receiver. Google,
which uses far better imagery than the Landsat, too has an offset
between the map data and the satellite images.


--
  Lars Aronsson ([hidden email])
  Aronsson Datateknik - http://aronsson.se

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

Re: GPS track oddities

Jonathan McDowell
In reply to this post by Nick Whitelegg-2
On Fri, Oct 14, 2005 at 06:29:01PM +0100, Nick Whitelegg wrote:
> Are you zoomed in too far? Landsat is only low resolution data so if
> you zoom in too far it becomes very indistinct. It's better for
> viewing say 5km x 5km areas. The main purpose for incorporating
> Landsat into osm-editor is for people to estimate polygons such as
> lakes etc, though it also provides interesting browsing over large
> areas.

I'm pretty sure the dark line on the Landsat data that's parallel to the
track but above it is the main road that the track is part of; while the
resolution of the Landsat data is low I can't think what else such a
dark line would be in this area.

J.

--
   Yes DM.  I'll just send my Iron Golem after those rust monsters...

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

Re: GPS track oddities

Nick Whitelegg

> I'm pretty sure the dark line on the Landsat data that's parallel to the
> track but above it is the main road that the track is part of; while the
> resolution of the Landsat data is low I can't think what else such a
> dark line would be in this area.
>

I have found the Landsat grabbed from osm-editor to be reasonably accurate (to
within about the width of a dual carriageway); any divergences appear to be
down to the GPS losing its signal rather than anything else. About 6 months
ago (see list archives circa April) I got some funny results but they
mysteriously disappeared just like that, so it looks like the Landsat server
was reconfigured...

e.g. on
http://www.free-map.org.uk/snapshot8.png
the coast path follows the coast quite nicely, and Hamble Station is at its
correct location at the intersection of the road and railway.

Nick



_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

RE: GPS track oddities

ANDY ROBINSON-2
In reply to this post by Jonathan McDowell
Just for info, I checked the location of your roundabout centre on Mapsource
which I have found to pretty reliable with respect to positioning in
relation to the Garmin GPS for which the maps are designed. It computed from
the map as N52.64531 E1.27638. Near as damn it the same on Google Earth.

The Latsat images are way off though. I don't think I have drawn a single
segment line (and it's been a lot over the last week or so) that has been
remotely related to the underlying map. I've travelled a lot of route where
Alex (Moriati) has travelled and out GPS plots are within the expected gps
errors. In fact we generally have very good correlation.

I assume the Lansat image frames are similar to aerial photo frames? These
need to be stretched to fit whatever projection/transformation is used so
may be that's where the errors comes from too. May not be anything related
to OSM at all.

Of course we could all be using the wrong datum :^) ...WGS 84 in the case of
OSM. If you ever find you are about 100m east of where you think you are you
can bet there is a translation to the wrong datum figuring somewhere (eg WGS
72).

Finally there is continental drift. We (Europe that is) are heading east
away from the USA by 2.5cm per year they say.... so how long will it be
before OSM needs to be "corrected" I wonder :)

The bottom line is that precision is all relative. Provided all the roads
meet the roundabout I hope everyone is happy. I'm sure that OSM will never
reach the level where it becomes a safety critical application.

Andy Robinson
[hidden email]

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Jonathan McDowell
Sent: 14 October 2005 18:13
To: [hidden email]
Subject: [Openstreetmap] GPS track oddities

I've been trying to produce some tracks around Norwich, but have been
having problems with my tracks appearing out compared to the Landsat
maps. At first I thought this was due to my cheap USB GPS receiver, but
I tried with a Magellan hand held device as well with similar problems.

So I tried something else. I used Streetmap to look up co-ordinates of a
roundabout (52.645225 1.276237) and a bit further up that road
(52.645992 1.273559).

Looking at the points with Google Maps shows them as you'd expect:

http://maps.google.co.uk/maps?q=52.645225+1.276237&iwloc=A&hl=en
http://maps.google.co.uk/maps?q=52.645992+1.273559&iwloc=A&hl=en

So I created a GPX file with these 2 points in it:

<gpx version="1.0" creator="A quick Perl hack"
        xmlns="http://www.topografix.com/GPX/1/0">
<trk>
<trkseg>
<trkpt lat="52.645225" lon="1.276237">
<time>2005-10-14T17:40:02Z</time>
</trkpt>
<trkpt lat="52.645992" lon="1.273559">
<time>2005-10-14T17:41:02Z</time>
</trkpt>
</trkseg>
</trk>
</gpx>

and fed this into osm-editor. Tell it to pull the Landsat data and I get
an image like:

http://the.earth.li/~noodles/osm-wrong.png

(248k)

This seems to me that I'm not going insane and that there's something
wrong with either the LandSat data or the way osm-editor and OSM deal
with it. Can anyone shed any light on this?

J.

--
    Love is an attraction to a     |  .''`.  Debian GNU/Linux Developer
  perfectly normal person you've   | : :' :  Happy to accept PGP signed
  temporarily mistaken for a god.  | `. `'   or encrypted mail - RSA +
                                   |   `-    DSA keys on the keyservers.

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap



_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap