[OSM-talk] augmenting contour data with gps track logs

Robin Paulson robin.paulson at gmail.com
Mon Dec 28 21:40:15 GMT 2009


2009/12/28 Frederik Ramm <frederik at remote.org>:
> Pure GPS elevation is worthless due to bad precision (for technical reasons,
> Z measurement is much worse than X/Y measurement in GPSes). Some GPSes have

ok, i didn't realise that. what levels of bad precision are we talking
about; i assume it varies form model to model; how does it compare to
the precision of srtm?

> We would need a mechanism that takes the spot heights measured with your GPS
> and somehow weaves them into this (or a finer-resolution) bitmap, *or* go
> for a true 3D surface model of the earth. Both will require different kinds
> of editors than we have now.

i would envisage it being used in favour of the srtm data (which is
very coarse, as you point out, hence my interest in replacing it)

>
> Simply adding altitude information to features we map will not do because
> most editors are unaware of that, and it is unclear what it means exactly.
> If you map a post box with altitude=100ft, and I later find that the post
> box is in fact on the other side of the road - does that mean that this side
> of the road is still 100ft, just the post box has to be moved, or does that
> mean that I have to move the altitude information together with the post
> box?

the idea was not to get people to manually tag the altitude of objects
(like post boxes), partly for the reasons you point out, partly
because it's slow.

my idea was to extract a series pf lat, lon, alt triplet from the gps
tracks, and use this to build a 'mesh' of the earth's surface, which
would be stored independently of the points/ways which we create at
the moment. the implicit assumption being that wherever the post box
or any other object is placed on the globe, it takes it's height from
the underlying mesh. from this mesh, we can then generate contours, 3d
models, etc, etc as we do at present with the srtm data

>
> In my eyes, a separate project "OpenTerrainModel" or something like that is
> called for, the results of which could seamlessly replace our usage of SRTM
> data today.

exactly

is there an api for accessing the gpx files en masse? or a gpx.planet download?




More information about the talk mailing list