Moving/Realigning large urban areas

abc26324 · 2019-10-22T21:21:01.000Z

Hello all,

I have been looking to try and get some accurate coordinates to which all else can be aligned, for example map layers such as Bing, which can often be out of sync by several meters.

I’ve got hold of the Ordnance Survey survey point data (Open Source) from here: https://www.ordnancesurvey.co.uk/gps/legacy-control-information/triangulation-stations which is, as far as I can tell, accurate down to around 10cm.

Converting from OSBG36 to ETRS89 the data can then be displayed in JOSM. As such, it has become apparent that many areas are off by several meters.

Is there any recommendations or precedents for moving large amounts of ways/nodes to align with this data (say several roads and surrounding areas at once)? I am of course very wary of doing this!

Further, in general, are there any efforts to try and match mapping to specific points of object we know to be, essentially, accurate. That is to say, if we know, say, a church tower centre point to be at a certain coordinate to cm accuracy, we should take these reference points and work from there?

And of course please let me know if my above thinking is out of whack!

Thanks in advance!

SK53 · 2019-10-23T12:08:25.000Z

Using OSGB36 will inevitably introduce further errors as it is an entirely outmoded co-ordinate system for anyone wanting highly accurate locations. Also depending on the conversion technique you used the RMS errors in conversion may well be in the order of 4-5 metres (i.e., if you do a straight conversion using proj4 utilities in ogr2ogr, postgis etc which I imagine is what JOSM uses).

So first of all you need to use the latest OSGB standard which I think is 2015. There’s an earlier one from 2002. For accurate conversion you need to apply specific corrections provided as a file by the OS. They have some detailed web pages about the conversions.

In general I would not advise trying to do such large scale rejigging as it’s pretty hard to get right: OSM objects will vary rarely be all created on the basis of the same underlying area. An accuracy of within 5m is what we can aspire to, trying to be more precise is not particularly worthwhile at present. In general when I’ve done this I’ve done it on smallish areas at a time and NOT by mass movement of nodes. You need to choose an area where the difference in position does not create obvious anomalies in surrounding areas which is what usually happens.

The StreetView base layer was created using the OSGB-2002 projection with the detailed correction, so this is a reasonable way to check for misalignment (mainly of larger roads, not buildings). The road network in OS Local Roads can also be used as a cross-check. At some point we ought to highlight roads which seriously diverge from this network (the techniques are not that complicated): there are many minor roads in Wales & the Welsh Borders which are significantly more than 10m away from true.

kucai · 2019-10-23T13:59:58.000Z

In addition, having aligning imageries to precise RTK fixed surveyed positions, different imagery providers have different distortion despite already corrected to the surveyed point. In my region, Maxar Premium has newer imagery, but Bing is much more beautiful and sharper. I just mix and match, and already made peace with “good enough” accuracy.