[Openstreetmap] Re: How do we handle large amount of GPS points?

[Petter Reinholdtsen]

[Joachim Mæland]
> I'm Petter's Linux user, holding ~7000k GPS points and new on this list... 

Welcome. :)

I'm glad to see you on the list. :)

> Go for smaller sections. There is no way you can predict accuracy
> and layout of the points, based on date only. Unless you run the
> track in a systematic manner.

My idea is to split between "tracks", sorted on date and with
approximately 1 month between each split point.  I define a track as a
set of GPS points taken in a row without any missing points.  At the
moment, I create a new track if there is more then 3 seconds between
one point at the next, and this is based on 1 second sampling rate.

I believe it is easier to find the end of tracks if I sort on date.
The alternative is to try to find all points within a region, and
follow each track at the edges to the end of the track to include it
in the dump.  I do not want to try to do that.

> IMHO: There are no NMEA information available that can be used for
> trusting GPS points blindly...

Oh, I am not talking about trusting them blindly, I'm talking about
reducing the inaccuracy.  Check out GPS Toolkit and the september
issue of Linux Journal for an idea of what I am talking about.

> My mandatory tools logs GPS positions, timestamp from localhost and fix. 
> Limited resources have stopped further enhancements, but my wishlist 
> includes: "GPS timestamp, DPOP, HDOP, VDOP and # of active satellites.

My script collect HDOP and #satellites, but lack the others.  I plan
to add GPS timestamp (hm, why is this timestamp several seconds off?),
and perhaps VDOP.  Not sure what DPOP is.

> The overhead involved in keeping this information is neglectable,
> compared to usefullness in special areas, where the accuracy of your
> dataset is questionable. You can use this information to disregard
> the points, not to trust them.

I'm currently looking at OpenGIS SQL standard, and suspect it is a
good idea to store the points using this framework.  I'll look at
PostGIS next.  PostGIS is an implementation of OpenGIS SQL for
PostgreSQL.  qgis support PostGIS.  Not sure what other tools support
OpenGIS.