[OSM-talk] osmfilter, Calculating Speed, help appreciated

[Joerg Ostertag (OSM Munich/Germany)]

On Thursday 07 September 2006 14:48, you wrote:
> > Example Speed:
> > If you have the speed driving along a road you can get a pretty good
> > guess which highway= tag this part of the street would be appropriate to
> > tag. Some of you might think; that is no problem, we have the timestamps
> > and the
> > position, so we can calculate the speed later too. But the timestamps in
> > your
> > gpx log have a precision of 1 second. So the speed might vary by almost
> > factor 2 (between 1 sec and 1.9sec) which for example at 90Km/h could
> > also be
> > 180Km/h which is not really precise.
>
> This is a bogus argument. If you have track points at ~1sec spacings, then
> it's fairly easy to use a moving average over 5 or 6 points to give a more

My tests told me 5-6 points are not accurate enough. This would still be 10% 
inaccuracy.

> accurate speed. If the points are at >>1sec spacing then the aliasing of
> the time value isn't important.
>
> As an aside, I've no idea how the compression algorithm actually works, but
> if it were me writing it I'd do something like:

If you can test this against your tracks would be great. osmfilter already has 
a floating speed calculation (exactly because of the inacuracy) which 
calculates the speed by always floating over 20 track points (last 10 and 
next 10). So please test this and tell me if the speed is accurate enough for 
it.
If this works we/someone can go one step further and add a classificator which 
takes the speed of a track segment and tags it according to the average 
speed.

Any help in this part is appreciated. Since my time is pretty limited with all 
the other stuff I'm trying to do.

Things which would be needed for osmfilter.pl are:
 - small test files for every input method(nmea,gpx,gps,mdb,gdb, ...)
 - small test Files for every option (--filter-clew,--osm, ...)
 - maybe a small shell script which 
    - tests that the filter is sill reading all formats 
    - and that it really filters what it should.
 - New filters to do more fancy stuff.
 - Reporting 
   - Bugs: Best would be with a fix ;-) 
           OK OK with an example triggering the Bug might be OK too.
   - Wishes: Best would be with a suggested patch;-)
   - New Working filters: As patch or just check it into svn directly ;-)

-

Joerg