[OSM-newbies] gps noise in osm database

[Paul Johnson]

Richard Weait wrote:

> On Sat, Mar 6, 2010 at 4:08 AM, Paul Johnson <baloo at ursamundi.org> wrote:
>> Jonathan Bennett wrote:
>>
>>> No. GPS traces are just evidence, or a guide, and shouldn't be
>>> interpreted literally. If they look messy, it's OK, since they're only
>>> used for tracing. The amount of work needed to clean up the traces would
>>> be massive (it would need to be done by hand), and we'd gain very little
>>> -- it doesn't generally affect the accuracy of mapping.
>>
>> Surely there's an automagical way to suppress bird's nests.
>
> gpsbabel filters simplify and discard sort of do this, but not exactly.
>
> gpsbabel position filter will allow you to remove points within
> $distance of a specific location, so that's good for removing a cloud
> around your customary home location.

So the problem's half solved.  Now we just need to figure out how to
detect a ratsnest location.

> I don't see a filter that both leaves 1/second points in place and
> reduces clouds from stationary receivers.

Upon detecting nonlinear data in a nonsensical fashion in a single
track, remove superfluous data points.

> Perhaps you'll be inspired by filter_skeleton.c to write an ideal
> anti-birdsnest filter for gpsbabel?

I would be, if it weren't for the fact I'm a civil engineering
technologist, not a computer scientist.  Sure, I /could/ do it, if
you want to wait forever for something that would almost certainly
be mistaken for unreliable crap. ;o)

> Also, the imperfections of gpx track files do remind us that the track
> file is not a complete survey.

I'm not convinced that's a valid excuse for not doing (basic) data
validation, particularly when it's been suggested and agreed upon
that it could be done.