[HOT] Question about background image alignment

[Dan S]

2014-08-13 17:20 GMT+01:00 Erik Walthinsen <omega at omegacs.net>:
> On 08/13/2014 08:44 AM, Dan S wrote:
>>
>> The answer is no, when you first load an imagery layer into JOSM it
>> does not offset it, and there may be a systematic misalignment
>> relative to GPS coordinates.
>
>
> OK, I guess I wasn't quite precise enough in my query:
>
> What I'm seeing is that much of the data in e.g. Monrovia is offset by 7.5m
> relative to the Bing imagery.  That means that either Bing is misaligned, or
> the imagery being used by many of the people entering data is misaligned (or
> both, most likely).

Those people *may* have used bing but corrected the imagery offset
before tracing, by aligning the imagery to some GPS. I don't know, I
haven't looked, but that would be the most optimistic state of
affairs! It's the best way to work.

> Potlatch2, iD, and JOSM all show the same offset vs. Bing, so we can
> eliminate *those* editors as the issue.
>
> Even worse, the particular ways I'm looking at (e.g. way 296966938) claims a
> source of Bing, which means that either they *aren't* using Bing imagery yet
> claiming to do so, or they are using some other mapping program that's
> faulty.

OR they are indeed using bing and are offsetting the imagery before
doing their work.

If you look at the history for the changeset in which your way was added:
http://www.openstreetmap.org/changeset/24645722
you see "created_by JOSM/1.5 (7347 en)" so we know it was JOSM.

I suspect they manually aligned bing before tracing.

(You could even ask them, I guess!)

You're right that they may also have used the alternate imagery source
that you mention, and forgotten to fix the source tag. But the
solution isn't to pick which imagery is canonical, the solution is to
work out how much offset the imagery in your area needs, in order to
align it to GPS.

When I look at the GPS tracks in the area you mention, I see pretty
good alignment between the road data and GPS. (I like to use the
rainbow GPS tiles to view this:
<tms:http://{switch:a,b,c}.gps-tile.openstreetmap.org/lines/{zoom}/{x}/{y}.png>.)
There are not many tracks in the area, but the data doesn't look to me
to be systematically misaligned, it looks to be probably correct.

Hope this helps!
Best
Dan


> Another example would be node 1102111387 which I corrected a few days ago.
> It was last edited in 2012 and also claims a Bing source, yet was offset the
> same ~7.5m east relative to what JOSM shows.
>
> A couple days ago I agreed to a license for an alternate imagery set, but
> the TMS link didn't work in JOSM, so I dropped it.  I can't find it anymore
> on the hotosm page, so I can't determine if it's the original source for
> this data.
>
> Whether or not Bing or this other imagery is more correct isn't necessarily
> the issue right now.  The problem is that data is being entered with one,
> then corrected against another, then maybe realigned back to the original
> offset.  Where one region is adjusted, it screws up everything on its edit
> boundary.
>
> *One* of them should be deemed "canonical" in the short term, and the Bing
> imagery seems to be the best candidate because it's the default in all the
> editors.
>> More generally, you can use the Imagery Offset Database to find out if
>> other people have suggested an offset for the area+imagery you're
>> working with: http://wiki.openstreetmap.org/wiki/Imagery_Offset_Database
>> In particular, check out the JOSM plugin which checks the database for
>> you very simply.
>
> There are no offsets for the Bing imagery anywhere in the area....