OpenStreetMap

I've been making quite a few changes to OSM from Bing. I moved a lot of roads slightly and wanted to check that Bing agrees with GPS. I drove from Hove down The Drive to the A259 coast road and along to Newhaven via Ovingdean and Saltdean. From Newhaven I went inland along the Lewes Road/Kingston Road to the A27. I drove to the A283 in Shoreham-by-Sea taking in my old house where I've added missing residential walks and garage compound driveways. I checked one driveway and one walk with GPS. I've corrected the topology of the roads and fly-overs at the entrance to Brighton Marina.

Location: Aldrington, Hove, Brighton and Hove, England, BN3 5HJ, United Kingdom

Discussion

Comment from Pink Duck on 5 May 2011 at 14:36

Personally I prefer to align the Bing aerial to the OS_OpenData_StreetView tiles based on buildings and junction central points, on the basis that the OS tiles after conversion to WGS84 are still likely the highest precision available in terms of allowable use.

Comment from chillly on 7 May 2011 at 10:14

The OS Street View layer available in OSM seems to be consistently some metres further south than multiple GPS traces. Our OS layers are only as good as the way they were reprojected. Bing alignment however seems to vary from place to place, sometimes very good, sometimes not so good and if there is a discrepancy it varies in direction.

A single GPS trace is not enough to correct alignment. Junctions with multiple crossing GPS tracks seem a good way to align the images. Using OS StreetView buildings for alignment seems very suspect. OS StreetView buildings are simplified and how can you use a GPS to align to a building? If you do manage to walk around the outside edge of a building to gather a GPS trace, the building will give you reflection errors.

Comment from ajrobb on 7 May 2011 at 10:49

I have collected some GPS traces to check some of the shifts. For instance, there is an open oval at the centre of Saltdean that was was previously about 20 metres off from Bing and OS. I have driven around it to check my adjusted positions with GPS and was pleased with the correlation. Walking in urban canyons gives random GPS but walking down the middle of suburban roads (you often have to anyway because cars are parked on the footpaths) seems consistent. A lot of my changes to roads have been to correct shape and layout as much as position.

Log in to leave a comment