[Openstreetmap] RoadMap map generator / editor

Andy Robinson Andy_J_Robinson at blueyonder.co.uk
Mon Oct 17 11:53:00 BST 2005


It might be possible to do this on a context basis using the timestamps for
the track. It is the timestamp that defines other aspects of the track,
including direction of route. If each track can be configured into a
"signature" using its time and positional data then it would help with the
context. For example it would help distinguish between a footpath and a
highway where the distance travelled in the time varies considerably.
Similarly, to use your example, a one mile signature for the A1 might be
very different from a one mile for Swanland Road. I assume also that when
each track point is written to the database it can have an associated track
ID assigned to it, perhaps the ID might be limited to a certain time or
length of track to break up large tracks and simplify a signature.

The one area where I know my own tracks completely mess up the prospect of
automatic line generation is when my gps emerges around a bend out of or
under tree cover. Very rarely will it pick up the correct alignment unless I
stop and let it catch up properly.

With respect to changes in road design with time I have already experience
of this problem on OSM in central Birmingham where the area around the Bull
Ring has recently be "redrawn" by the planners. The presence of old tracks
in the editor makes discerning a new and more reliable track more difficult.
Having the most recent tracks (or perhaps your own tracks) in another colour
(eg yellow) in the editor would help see the wood for the trees.
For automatic change of street configuration this could again be done on a
time basis. The map does not get changed (unless it's done manually) until
the critical mass of more recent data outweighs the older data. I appreciate
though that this will be nigh on impossible to achieve when changes in
configuration are small and would result in a considerable lag unless the
newer data is heavily weighted.

On the whole merging issue I'm going to spend some time with my spreadsheets
on numerical analysis of lines and nodes (where line cross) using some gps
data I'll collect from a set of street blocks and see what pops out. Anyone
else actively working on this aspect?

Andy Robinson

-----Original Message-----
From: openstreetmap-bounces at vr.ucl.ac.uk
[mailto:openstreetmap-bounces at vr.ucl.ac.uk] On Behalf Of Simon Hewison
Sent: 17 October 2005 11:18
To: openstreetmap at vr.ucl.ac.uk
Subject: Re: [Openstreetmap] RoadMap map generator / editor

Tom Carden wrote:
> So the ideas I'm after are thoughts people have on merging an existing 
> road with a GPS track.  Or in OSM's case, an existing road with hundreds 
> of GPS tracks of that road.  When do we stop refining the shape?  How do 
> we take noise into account?  How much priority do we give to the most 
> recent data?  ( e.g. if a new roundabout/intersection is added, and we 
> have 100 GPS tracks which say there isn't one, but 1 GPS track for today 
> which says there is - can that be automatically merged?)

Good example here. At
Lat:51.6992479965827 Lon:-0.2262137092125

(A1(M) south, near Potters Bar), there's a parallel road that is within 
a typical GPS error of the main road. It follows the motorway for a long 
way.

How would the road merging code know that the track was taken while 
driving south along the A1(M) and not the parallel Swanland Road? The 
only way I can think is that it would have to take context into account 
(the only sane way to get onto the A1(M) southbound is via a slip road, 
and Swanland Road starts where there is no specific interchange with the 
A1(M))

You certainly wouldn't want the code to swing Swanland Road towards the 
A1(M), or vice-versa.

-- 
Simon Hewison

_______________________________________________
Openstreetmap mailing list
Openstreetmap at vr.ucl.ac.uk
http://bat.vr.ucl.ac.uk/cgi-bin/mailman/listinfo/openstreetmap






More information about the talk mailing list