[talk-au] Misaligned streets OSM or Bing wrong? - use survey mark?

[Ben Johnson]

Hi Russell,

Welcome aboard. Just my thoughts...

For a newbie, you've made some very astute observations about the accuracy of the aerial imagery, so thanks for giving it the degree of thought - because many others just dive in and start tracing without understanding these subtleties.

You're absolutely correct. Its quite normal for Bing (and others) to be off by 10m or so, but in other areas it's very accurate. It differs all over the place.

Whatever editing tool you're using (Potlatch, JOSM, etc..), should give you the ability to align your background imagery. Just be aware the further you move away from the area you have aligned, the more likely the alignment will need another adjustment.

JOSM allows you "bookmark" such alignments. I have a bookmark for my home town. But the danger of bookmarking your aerial alignment is that if Bing updates their imagery, your bookmarked alignment will probably no longer apply. Just something to keep in mind.

Your known survey mark is definitely an excellent start. You can make a node in OSM to those exact coordinates, then align your background imagery so the location of the survey mark in Bing aligns to your OSM node.

I would not rely on other people's OSM edits on which to base your alignment. You don't know how accurate or inaccurate they are. Your survey mark is the best way, and the GPS traces.

Are you aware you can upload and share your GPS traces (in your profile section), and you can view all the uploaded traces while editing? This is great for aligning the background image. Depending where you live, you might have some major roads in your area which already have lots of traces. If so, that's great - try and find a couple of busy intersections to align and you should get a good alignment.

Otherwise you can try walking around a block several times, or around a park, or around a roundabout many many times (if you see men in white coming to get you... Run!) Another method is to use a fixed point - and approach it from different angles like cross hairs to a target.

As far as GPX averaging goes, I'm personally not familiar with anything that does this, but this list has some very talented people who can help or point you in the right direction. I find software averaging is not necessary because if you look at the traces visually you'll immediately get a very good idea from their density when they're all presented together (eg if you look at the traces on a busy dual carriageway motorway you'll very easily see an "average" for each carriageway). In built up areas you just need to be mindful of whether traces originate from cyclists, motorists, or pedestrians... and apply a degree of commonsense. I'm not sure if software can possibly distinguish between a set of traces from a roadway, and another set of traces from an adjacent walkway / cycleway - in such cases an average is meaningless.

As far as tweaking other people's edits... if you're confident with your accuracy, I'd just do it. But if you want to take a more cautious approach you can contact the original editor and ask whether they mind you tweaking - or at least ask what method they used to position their nodes and let them know you can make it more accurate for them. Keep in mind there are many "inactive" users who move onto other hobbies after a while - so if you don't hear back from somebody after a reasonable time, consider it fair game to change it.

Again, well done for going through this thought process, and welcome!

BJ



Sent from my iPhone

On 01/09/2012, at 8:11, Russell Edwards <russell2pi at gmail.com> wrote:

> Hello -- brand new user here, please be patient :).
> 
> I am trying to improve OSM in my home town. I notice that many/most OSM ways are approx 5-10 m west of where Bing has them.
> 
> Most either have nearmap or nothing as a source. There are no GPS tracks to download. I could make some but they're usually in (random) error by about the same amount anyway.
> 
> I tried to check against the coordinates given for a survey mark in town, through http://services.land.vic.gov.au/maps/lassi.jsp (with conversion to lat/lon with http://www.gracode.com/mapgrids.php ) ... this also had the Bing map out by about 10 m... but southward!
> 
> Help! Three inconsistent systems - Bing, OSM and the survey mark. Which if any should I trust at the 1-m level? 
> 
> Is there a tool for averaging GPS tracks? There are roads I have run along dozens of times with a GPS track for every run. Maybe with averaging it could get to the 1m level of accuracy.
> 
> (Yeah yeah, don't sweat the small stuff... I just can't stand seeing streets running through people's front gardens...)
> 
> Thanks in advance
> 
> Russell
> 
> _______________________________________________
> Talk-au mailing list
> Talk-au at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-au
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-au/attachments/20120901/9441a223/attachment.html>