[talk-au] Mass realignment of roads? (was Mapping things by importance)

[Delta Foxtrot]

--- On Fri, 12/6/09, James Livingston <doctau at mac.com> wrote:
> Unless you have a known point of reference, clearly visible
> in the imagery and placed via something accurate (i.e. gps),
> they can be out a fair bit.

You can also get into datum issues, OSM uses WGS84, which differs from DGA94 by about 1m at present (my best guess, still trying to find a surveyor to confirm), however as time goes on this increases by about 7cm per year due to the Australian continental drift.

However that isn't the biggest source of error, most consumer grade GPSr units are only qualified accurate to about 10m, although this can be minimised by doing an average of location data over a 24hr period, although this is fairly time consuming obviously.

Until we have dual frequency receivers to eliminate sources of error with atmospheric conditions 10m accuracy should be acceptable for most purposes, unless you are a surveyor of course.

> I've come to the realisation that every road (and
> everything that was placed in reference to the roads) in the
> suburb where I live is out by about 8-10m. Much of
> Brisbane's road system seems to have been traced off the
> Yahoo imagery, apparently without aligning it to fixed
> points before hand.

8-10m isn't that much of an error to be honest. There really isn't any such thing as perfect data if you are using consumer grade equipment of any sort, things will improve over the next 10 to 20 years with other constellations of GPS like sats starting to broadcast, but until then 10m is as good as it gets for the most part for most things unless you spend a lot of time getting very very accurate positioning.

> When I've been going around recently, I started noticing
> that POIs on the north-east corner of an intersection looked
> like they were on the SW corner. Obviously the best solution
> is to go out and capture all the roads with my GPS, but I'm
> wondering what to do in the mean time.

That can happen at diff zoom levels because of the x,y offset position that they are plotted from also.

> If I leave the roads where they are, the data I'm putting
> in will look wrong as it's not consistent with where the
> roads have been placed. Moving just the intersection will
> make the road appear to bend in places it doesn't, and would
> be a fairly arbitrary thing. I could go around and move
> every road in the suburb the same distance, but that'd be a
> fair bit of work, and I'd probably screw something up.

IMHO if you have a better source of information it should be used, and over time errors should be corrected by others till things do look right, since seeing mistakes motivates people to fix them and in some cases to go on and do other things too.