OpenStreetMap
discrepancy between high resolution bing satallite maps and OSM
Posted by DocSeppi on 12 April 2011 in English.Hi All,
I wanted to edit parts of my home region with the help of the new bing satellite
maps and came across that there are slight differences between the roads already
tagged in OSM and the satellite pictures.
In particular I am speaking about this region:
http://www.openstreetmap.org/?lat=47.4661&lon=16.5705&zoom=13&layers=M
I'm now wondering what data is more correct: The one OSM or Bing Satellites?
If you think the satellite images are accurate I will edit the courses of the roads.
Thanks for your answer!
DS
Discussion
Comment from Todeskuh on 12 April 2011 at 15:22
Great, I already had the same thought.
When I compare OSM and satellite data the latter is ofter quite a bit off-track when it comes to GPS traces. The difference is always small, maybe less than 10 meters, but it's still a bit annoying.
Comment from Vincent de Phily on 12 April 2011 at 15:39
It's often hard to know which source is most trustworthy. Checking the "source" tag (if present) of various objetcs in the area might give you a hint. If *every object* in the area is missaligned (compared to bing) by the same amount and direction, that's another hint. If you can, place your gps in an easily-recognized open spot for a long while (even better: at a few hours or days' interval) so that you have a trustworthy gps coordinate that can be matched on an aerial photo.
Once you have a source you trust, just change the offset of the bing layer so that it is aligned correctly. And store that information as a point in the map, or on the wiki (not sure what the best practice is).
Comment from SK53 on 12 April 2011 at 16:01
In your case I think the answer is straightforward.
Some roads are both closely aligned with GPS traces & Bing imagery. Others are not: a quick inspection of some of these roads showed them to have source=plan:at. I believe these are imported data with a lower level of accuracy than OSM data mapped on the ground.
The best approach is to survey these roads using a GPS, you can then check the alignment of the Bing imagery and have the best of both worlds: a recent ground survey and the extra information from aerial imagery.
Comment from AndrewHutch on 12 April 2011 at 18:03
I've found it to be a bit of both. Quite a lot of GPS data is good to around 20m so sometimes the Bing data is better. Some of the Bing data does seem quite old though. I'm leaning towards tracing the Bing map, then surveying to confirm street names etc, which also gives a GPS trace, and an opportunity to check the printed map for errors.
Comment from ConsEbt on 12 April 2011 at 20:42
There is a corresponding note/section in the wiki http://wiki.openstreetmap.org/wiki/Bing#Precision
I've read through a few of the comments and what I remember is
a) have a look at major roads and prominent points to check the miss alignment of the BING images
b) missalignment can be different in every zoom level!
c) in potlache 2 the BING images can me moved by holding down space and move the background (use major points in the area to check)
Comment from Rovastar on 12 April 2011 at 23:33
To be honest I just use Bing imagery now. GPS traces seems to be so off (although I only use a mobile phone)
Also I have a yet to see a decent road with any curves mapped with any detail. The bing maps you can get to precision accuracy for the formation of the road/building/etc (I use the middle of the road/road marking for better accuracy and easier to map some of the main road/motorways are shockingly bad) the alignment may be off (I am sure all GPS as off anyway to some degree and they never agree) but the formation of the way is always going to be accurate. I don't see how you can get the formation of the ways any more accurate than sat imagery.
I only reason I would not use Bing is for old imagery and where things are mapped already in open spaces.
I have also realigned some imagery with bing and it completed the UK missing streets data in areas I haven't even visited before. They were off by more the agreed tolerances.
Comment from Rovastar on 12 April 2011 at 23:41
Also I thought the modern way to use Bing imagery now and less GPS traces. That is what most seem to be doing. Otherwise it is very tricky to map golf courses for example. Do people really work around all groups of trees and bunkers, etc? Surely not anymore.
Comment from Vincent de Phily on 13 April 2011 at 09:03
@Rovastar: The difference in accuracy between a phone and a special-purpose GPS is quite big. I even switched recently from a garmin eTrex to a GPSmap62 (I lost the previous device), and the newer model is slightly better still. Nobody claims that GPS traces are perfect, but they're invaluable even if you've got bing imagery in your area, to provide proper alignment for the later.
Comment from Rovastar on 13 April 2011 at 10:08
@Vincent de Phily
True I expected a dedicated device to be better than a phone. But between devices...I don't get it so your new device better or just different?? how do you know? Do you then go and remap to the new standards? Are you going to walk down the centre of the road? I thought GPS traces were now used to get a rough idea of where you where and then use bing imagery.
Anyway I don't think I will change and a new generation of people is coming in that will just use bing imagery. Mostly it sounds like FUD the pro-GPS arguments not to use it.
Comment from Vincent de Phily on 13 April 2011 at 13:32
@rovastar
the new device has a better (but bulkyer) antenna, and probably some software tuning. In practice, the difference is quite small but it is noticable : for example this one finds the satellite inside my house near the window, while the older one did not. I had the two devices available one day and tested them side by side, the new one reported a slightly better accuracy, and behaved much better when I took sharp turns in a narrow street with high buildings.
I will not remap anything I did before, the difference is not worth it. I did it just once for a particularly tricky street, but that's a rare case. If I can, I do walk (cycle) near the center of the street :) Or At least I try to remember on which side of the road I was walking.
I really do not want to downplay how great having bing imagery available is. I cannot thank Microsoft enough (for once...) for this authorisation. But bing imagery does need to be taken with a grain of salt, and a good GPS device can tell you exactly what seasoning is needed. If you look at 'source=' tags sometimes you'll see 'bing' and sometimes 'bing adjusted'. That's an important distinction.
disclaimers:
* I have only ever owned garmin GPSes, but I've seen online comparisons of iphone vs droid vs gpsmap, and the difference is big (bigger than between two garmin devices).
* bing hi-res isn't available in the main area I edit, although I used bing for a few other areas.
Comment from Anna_AG on 14 April 2011 at 00:51
BING imagery does seem to have offsets to GPS Traces around the world. I have edited a number of places with BING but cross references with my own multiple GPS traces taken over multiple periods ( important for accuracy apparently, as it changes / shifts during the day )
JOSM does allow you to adjust the BING imagery with an XY offset. If you ask JOSM to download the GPS traces ( highly recommended ) you can check the imagery aligns with the GPS traces, if not I align the imagery to the average trace - assuming a main road is close by with intersections, allowing northing and southing to be checked.
Assuming there are multiple traces of the GPS on roads, assume that the average of these traces is pretty much about as accurate as we can get right now.
Not all GPS devices are the same, modern units are more sensitive ( Garmin etc ) than previous versions and generally more accurate - 3-5 meters with larger antennas than that possible to fit into a mobile prone thus generally more accurate ( more signal / less noise ).
You can check the date of the pictures in Bing imagery here
http://mvexel.dev.openstreetmap.org/bing/