data explosion

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

data explosion

stevec-4
Hello everyone,

With much thanks to Saul we now have an arrangement to get lots of gps
tracks from ecourier. I successfully imported last weeks track of
300,000 points in and around london and its up for viewing.

These links show their data overlapping with tracks we've made:

http://www.openstreetmap.org/tmp/screen.jpg
http://www.openstreetmap.org/edit/applet.jsp?lat=51.52378&lon=-0.16888845&scale=23411.062

We can expect and additional similar number of points every week. In
central london this makes tracks quite dense, we may need to review how
they are displayed (eg as lines instead). Tom?

The script to turn their data in to gpx's is my first ruby script and is
in subversion. Its beta, please comment and / or change. I expect we
might re-import things if the script improves. When the other rubyists
here are happy, we can automate the weekly import of data.

For those interested, there are nearly 7 million points now in the
database. If anyone feels like doing some hacking, we have lots of
things TODO, please get in contact on the openstreetmap-dev list.

have fun,

SteveC [hidden email] http://www.fractalus.com/steve/

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

Re: data explosion

Lars Aronsson
SteveC wrote:
> With much thanks to Saul we now have an arrangement to get lots of gps
> tracks from ecourier. I successfully imported last weeks track of
> 300,000 points in and around london and its up for viewing.

That's great news!  Will the source data files be available for
download?  Can they be used outside of OSM?

> These links show their data overlapping with tracks we've made:
> http://www.openstreetmap.org/tmp/screen.jpg

Note how sparse the data are.  Just one or two dots per block.  
Seems they are using less than one-sample-per-second.  Just
looking at the dots (and not the straight lines between adjacent
dots), it will be very hard to determine where the roads are. This
could be compensated by interpolating the missing one-second
samples when the files are converted to GPX, but that would create
very many more trackpoints.  Rethinking OSM from dots to lines
would be the better approach.

> For those interested, there are nearly 7 million points now in the
> database.

Since the number of points depends entirely on the sample rate, I
think time-spent should be the measurement for track data volume.
How many hours of driving have we recorded?


--
  Lars Aronsson ([hidden email])
  Aronsson Datateknik - http://aronsson.se

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

Re: data explosion

stevec-4
* @ 06/07/05 01:18:13 AM [hidden email] wrote:
> SteveC wrote:
> > With much thanks to Saul we now have an arrangement to get lots of gps
> > tracks from ecourier. I successfully imported last weeks track of
> > 300,000 points in and around london and its up for viewing.
>
> That's great news!  Will the source data files be available for
> download?  Can they be used outside of OSM?

Yes, downloading gpx's is a planned feature.

> > These links show their data overlapping with tracks we've made:
> > http://www.openstreetmap.org/tmp/screen.jpg
>
> Note how sparse the data are.  Just one or two dots per block.  
> Seems they are using less than one-sample-per-second.  Just

They are once a second. Believe it or not, the screenie includes part of
a motorway (I think I can call the A/M4 Westway that).

> looking at the dots (and not the straight lines between adjacent
> dots), it will be very hard to determine where the roads are. This

It isn't hard to determine if you know the place! :-)

> How many hours of driving have we recorded?

I don't know. You could extract things via xmlrpc and find out, or wait
until I do some kind of database dump.

have fun,

SteveC [hidden email] http://www.fractalus.com/steve/

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap
Reply | Threaded
Open this post in threaded view
|

Re: data explosion

Glen Barnes
In reply to this post by Lars Aronsson
Lars Aronsson wrote:

>
>Note how sparse the data are.  Just one or two dots per block.  
>Seems they are using less than one-sample-per-second.  Just
>looking at the dots (and not the straight lines between adjacent
>dots), it will be very hard to determine where the roads are. This
>could be compensated by interpolating the missing one-second
>samples when the files are converted to GPX, but that would create
>very many more trackpoints.  Rethinking OSM from dots to lines
>would be the better approach.
>  
>
Surely after several weeks of data the dots will be filled in more? For
example if the same road was driven 20 times over several weeks then  
the dots would be a lot closer together making it easier to distinguish
roads.

Glen

_______________________________________________
Openstreetmap mailing list
[hidden email]
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap