[OSM-talk] relating trace points to gpx files

[Andrew Loughhead]

In general, Etienne, I fully support what you are saying. But I think
there are a couple of counter arguments:

 - Minor roads, residential cul-de-sacs, and similar, may only get
mapped as a deliberate activity, and so most likely the segments and
ways will be based on a single trace. In that case the trace should
probably be pretty clean to avoid confusion. 

 - One partial response to the privacy concern of Matt A, described by
Steve in his "long tail" thread, would involve control of what trace
points are uploaded. You might use as an exclude mask your suburb, or
some random shape encompassing your home, and prevent your traces
connecting to whatever locations you feel sensitive about. This is
really the same as stopping your receiver recording, and starting it
again later. 

 - The track point density control offered by Joerg's osm-filter script
I guess is another reason to edit gpx. 

Changing location information in gpx files is clearly a very bad idea,
as the question then arises as to what knowledge allowed location
changes. If a particular survey exercise produced poor data, the only
valid response is to collect more data. But cutting out parts of the
track is equivalent to stopping recording and starting again, which I
would do while driving or riding, if I could safely. As I can't, I think
such basic editing is reasonable. 

But in general, yes, I agree that there is no reason to get excited
about the odd bad point caused by multipath or poor sky view, or even by
an excursion into a car park. 

cheers
Andrew.



On Fri, 2006-07-14 at 15:32 +0100, Etienne wrote:
> One thing about this is that there will always be some "bad"
> trackpoints in OSM.  If you look at any city centre you'll see that
> the tracks can be very scattered.  It's quite common for a GPS to be
> "off" because it has picked up a reflected signal from a tall
> building.  
> 
> Over time, with many passes along the same street there will
> eventually be a strong _average_ scatter of the  trackpoints that will
> be be a good approximation to the actual line of a road, but many of
> the points will be quite a distance from the line and sometimes a very
> long distance.  Take a look at central London if you want to see
> streets with many thousands of trackpoints. 
> 
> As for whether you should cleanup/edit your tracklogs before uploading
> them - I've not seen any serious discussion here on that subject, but
> since the tracklogs are supposed to represent the raw data from which
> the refined maps are created I wonder if its actually a good thing to
> manually edit them before uploading?  
> 
> Etienne
> 
> On 7/14/06, Andrew Loughhead <andrew at incanberra.com.au> wrote:
>         On Fri, 2006-07-14 at 09:59 +0200, Erik Johansson wrote:
>         > On 7/14/06, Andrew Loughhead <andrew at incanberra.com.au>
>         wrote:
>         > > On Wed, 2006-07-12 at 10:57 +0200, Erik Johansson wrote: 
>         > > <--snip-->
>         > > > A far as I know (from reading the code)  when you delete
>         a GPX file,
>         > > > the points will be marked invisible. So they will
>         disappear from the
>         > > > rendered images. 
>         > [...]
>         > > I have been able to delete all my files and start loading
>         them
>         > > individually, checking each result in JOSM.
>         >
>         > Oh no... This thing about deleting GPX traces. I don't like
>         it, we 
>         > should never make people delete the files. I think it's
>         blasphemy to
>         > remove data from the traces db.
>         
>         I have finally found my problem file. As is traditional, it
>         was the last
>         file I looked at. I have deleted the 20 odd garbage track
>         points, and 
>         all data has been re-sent to OSM.  So far the applet display
>         still shows
>         the bad points.
>         
>         I could have found the bad data without deleting files from
>         OSM, if I
>         had tried a little harder (or to be honest, actually turned my
>         brain 
>         on). But it was an effective, if tedious, way of finding where
>         the bad
>         points came from.
>         
>         Andrew.
>         
>         
>         
>         
>         
>         >
>         > I also use josm and I do it like this:
>         > 1. load GPX file
>         > 2. download GPS traces for that area 
>         > 3. upload GPX file to OSM
>         > 4. edit
>         >
>         >
>         > > One question for the list: in JOSM it seems that if raw
>         gps mode is used
>         > > for an OSM download, that all points in view are treated
>         as being a 
>         > > single track, based on sequential time stamp?  The result
>         being that
>         > > relatively random straight segments appear just because
>         there happen to
>         > > be some track points in view?  The manual doesn't really
>         say. 
>         >
>         > The traces downloaded from the server are just dots on a map
>         nothing
>         > more. As far as I can remmeber they have no metadata.
>         >
>         > You should always create OSM data layers from a GPX file you
>         yourself 
>         > have created. Never from traces that is downloaded from the
>         server.
>         > And I'm not quite sure why it's not strictly verboten. Imi?
>         >
>         
>         
>         _______________________________________________
>         talk mailing list 
>         talk at openstreetmap.org
>         http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
>