[Talk-GB] Georeferencing / zeroing imagery

SK53 sk53.osm at gmail.com
Sat Sep 14 15:14:57 UTC 2019


Hi Edward,

In general the GPS rule is still the best way of doing it, we used to have
access to Strava heatmap which was very good, but no longer.

Other viable alternatives are:

   - OS OpenData road centrelines. Of course if you use a crude OSGB->WGS84
   based on OSGB36 this may be upto 5 m out anyway (maximum error of the
   straight conversion).
   - Open Data sets which have probably been very accurately located
   (Nottingham Streetlights, trees in Bristol, Birmingham etc). I presume
   councils use some kind of differential GPS to locate assets.
   - OS StreetView layer was rectified by Grant using OSGB-02 and is good
   to crosscheck across layers for shifts.
   - Lidar data from Enivornment Agency/SEPA etc doesn't suffer from
   parallax errors and can be useful for identifying shifts also.

The whole issue of conversions is vexed. There are a number of recent
diary/blog posts summarising the issues. Basically georectified imagery is
probably rectified to some OSGB standard here.

Imagery is always a mosaic you can drop off a consistent area at any point.
Nor are individual zoom layers always consistently georectified or even
from the same time. Old maps are generally poorly aligned compared to other
layers. The NLS 1:25k is usually OK. GSGS3906 in Northern Ireland is also
generally OK and can sometimes be enhanced by adding more warp points on
MapWarper. Something like 80% of ways mapped from GSGS and imagery is
within 5m of OSNI data.

In general the more effort put into aligning features the more pain when at
some later time one discovers that the alignment was out. I would only
really bother with gross misalignments (say >10m). Anything which is within
5m is unlikely to repay the effort of realignment. Always maintaining
topological relationships is however important. If one really wants greater
precision then OS sell quite a decent product.

There is scope with things like RTK to create a set of known locations with
known precise co-ordinates. People do this for drone imagery flown in
humanitarian situations.

HTH,

Jerry

On Sat, 14 Sep 2019 at 15:58, Edward Bainton <bainton.ete at gmail.com> wrote:

> Hi folks
>
> Another query from me that I imagine has been done to death elsewhere (so
> apologies), but I haven't found it on the wiki. The usual disclaimer that
> I'm not well up on the technical side of the map.
>
> I've read the wiki on Using Imagery
> <https://wiki.openstreetmap.org/wiki/Using_Imagery> and have understood
> that the best way of 'zeroing' the offset of imagery is to look for GPS
> traces, albeit even they are somewhat inaccurate.
>
> A few questions:
> - Over how wide an area does an offset obtained that way hold good?
> - Are the old OS maps better/worse/same as this system? Are they an
> alternative for zeroing imagery?
> - If I know the grid reference of somewhere (eg, Environment Agency puts a
> 10-fig reference on a plaque on their assets) is that any help?
>
> On the last point, this wiki page
> <https://wiki.openstreetmap.org/wiki/Key:ele> mentions OSGM02. The link
> has rotted so I searched their site which gives hits for OSGM15
> - Is 02 an outdated standard that I should update on the wiki?
>
> Thanks,
>
> Edward / eteb3
>
>
>
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20190914/de82c6dc/attachment.html>


More information about the Talk-GB mailing list