[OSM-talk] Retaggin of ways

Andy Robinson (blackadder) blackadderajr at googlemail.com
Thu Dec 6 10:13:02 GMT 2007


Jo wrote:
>Sent: 06 December 2007 8:13 AM
>To: Jon Burgess
>Cc: talk at openstreetmap.org
>Subject: Re: [OSM-talk] Retaggin of ways
>
>Jon Burgess wrote:
>> On Wed, 2007-12-05 at 17:22 +0100, Jo wrote:
>>
>>>>
>>>>
>>> It kind of bother me too that all the roads I touch show up with my name
>>> on them. (I'm touching a lot of them, since I'm using the Yahoo! imagery
>>> to line them up). Now the original editor is not visible anymore in
>>> Osmarender. I suppose it is know somewhere in the innards of the DB
>>> though and only the last person editing is shown for practical reasons.
>>>
>>>
>>
>> You need to be careful about the relative accuracy of different sources.
>> I believe that GPS traces are generally more accurate for road positions
>> than Yahoo! imagery (especially if there are several GPS traces covering
>> the same road).
>>
>> Adding roads based on Y! when there are no other traces is fine but
>> please don't re-align existing roads just to fit the Y! images.
>>
>> The only other obvious exception to this would be with roads consisting
>> of very widely spaced nodes. In these cases the roads could be enhanced
>> by adding a few more nodes for extra details visible in the Y! imagery.
>>
>> 	Jon
>>
>Quite often they are my own gpx-traces I'm 'improving'. I go in the
>field with the GPS, taking pictures (when by bicycle or on foot) or
>notes (when by bus). Then I enter those traces with josm, merging them
>with the extra info. Then I do a second/third pass to align it all with
>the Yahoo! imagery. Of course, while doing that it's very hard to resist
>the temptation to also change the ways the other local contributor put
>in in 2006. There are not many tracks to go by anyway. He didn't upload
>many and I don't upload all of mine either.
>
>The way I understood it, GPS-tracks can be off for about 10 meters. At
>least, when it's all aligned with the aerial pictures it will be more
>correct and proportional compared with the surrounding streets and
>features. Even if it's off a bit, it will be good enough for the time
>being.
>
>Jo

Don't be too over confident about the Yahoo imagery position. It's only as
good as the rectification process of the original data and we know of
examples where it's off by the kilometre rather than the metre. This type of
large error is rare but it's easy to find examples where the imagery differs
from good GPX traces by anything up to 30m. The point is it's not a fixed
discrepancy. Likewise NPE and Landast have their own errors too (Landsat
often in the order of +/-100m and NPE depends upon the stretch/warp of the
original maps scanned - getting better as the originals are corrected)

As you say, GPS accuracy might for a single trace be in the order of +/- 10m
if you have had good sight of the satellites. It might be a little better if
you had SBAS functioning (EGNOS/WASS) and it might be a bit worse if you
have poor satellite position or there were other external influences (eg
trees etc).

The point is though that realigning using more than one GPS trace as well as
the aerial imagery is I believe the way it should be. The aerial imagery
generally gives a much closer shape to the road system and the GPX traces
(if there are several) probably give the best reliability for overall
position of the street. So using both sources for areas with Yahoo imagery
and not relying upon any one is the way to go.

With respect to other peoples edits it is entirely appropriate that you
tweak other editor's contributions where you feel all the validating data
indicates a tweak is desirable. That's the whole point of a wiki map. But I
agree that realigning map data just to fit one source without reference to
any other is a bad thing which makes it so important to upload your gpx
traces to the server and leave them there for others to reference.

If you really do have reason to want to leave an object exactly as it is
then put a note= in there so that others can see and evaluate your
reasoning.

Cheers

Andy





More information about the talk mailing list