[OSM-dev] collection of gps tracks

[Adrian de Rivero]

Thanks for your replies.

2008/10/23 Andy Robinson (blackadder-lists) <ajrlists at googlemail.com>

> Patrick Kilian wrote:
> >Sent: 23 October 2008 12:30 PM
> >To: Shaun McDonald; OSM dev list
> >Subject: Re: [OSM-dev] collection of gps tracks
> >
> >Hi,
> >
> >>> In more detail:
> >>> I'm looking for an area with a high density of gps traces (collected
> >>> only by car).
> >>> Where a single road can be assign to several Tracks (10 and more).
> >>
> >> When we upload a GPS trace, there is no tagging as to the mode of
> >> transport. Thus finding this information will be difficult.
> >Traveling by car should have some features which can be detected
> >statistically. High speed, small slowdown when going uphill, rarely
> >passing over footways etc.
> >
> >
> >>> The Idea:
> >>> To extract informations of a collection of tracks that can't be
> >>> extracted by a single one.
> >>> As an example: You could extract the number of lines for a road and
> >>> tag it in OSM.
> >I had that idea myself, but didn't have time to actually try it.
> >
> >
> >> GPS traces are unlikely to be accurate enough to be able to get the
> >> number of lanes on the road.
> >If you have only a docent track? right. But what if you have 100 tracks?
> >Or 1000? It's going to need a health dose of advanced statistics so.
> >
>
> It doesn't work because each time I drive down the same lane the GPS will
> give position of that lane within a few metres but each trace on different
> days will not exactly overlap each other. In reality because of satellite
> position and numbers of satellites on view at any given point in time the
> positional accuracy will vary. As a result you cannot discern individual
> lanes from multiple tracks. The only way you reliably know the number of
> lanes is to record the number or to drive all lanes immediately after each
> other. the number of passes then = the number of lanes but would of course
> be very time consuming. Noting the number of lanes during survey is so much
> easier ;-)


I don't want to start to much of a discussion if it's possible or not. But
if your are interested here are two papers that talk about it.

Schroedl, S., Wagstaff, K., Rogers, S., Langley, P., and Wilson, C. 2004.
Mining GPS Traces for Map Refinement. Data Min. Knowl. Discov. 9, 1 (Jul.
2004)

Edelkamp, S. and Schrödl, S. 2003. Route planning and map inference with
global positioning traces. In Computer Science in Perspective, R. Klein, H.
Six, and L. Wegner, Eds. Lecture Notes In Computer Science. Springer-Verlag
New York, New York, NY

As Patrik mentions that with a large number of traces the overall error
should decrease.


>
>
> If a dense area of GPS data is needed, try Sutton Coldfield, UK where I
> have
> a huge number of tracks to and from uploaded. Its obvious which are by car
> as I generally don't cycle much above 30km/hr ;-)
>
>
I looked at Sutton Coldfield and it seems that it could work.
But I can't download the original gpx tracks.
The only possible way to extract the data would be with the api. The
boundingbox function is not what I am looking for, because I want the
original tracks (i.e. with timestamp) and it would require to split up the
the downloaded tracks what I would like to avoid.
Do you still have your tracks? And can I have them?

Anybody else is welcome to send me their tracks! But as I wrote in my first
mail, I need tracks for an area with multiple tracks for a single road and
they should be collected by car.

Cheers
Adrian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20081024/d0a7cf4f/attachment.html>