[HOT] Somalia import from 2011

Harry Wood mail at harrywood.co.uk
Mon Nov 5 17:27:28 GMT 2012


It's data like here in Markar we're talking about still right? http://osm.org/go/wq63Lv3C

I don't think we should *bulk* delete any data in Somalia, but if I was working on the map in this area I would have no hesitation in deleting data like this and re-adding road layouts back in from bing. 

The imported data might possibly be geolocated (by GPS) more accurately than bing imagery... possibly...  but it's clearly *very* inaccurate in terms of basic topological road layout. I think we need to be bold about fixing it up.

Harry



----- Original Message -----
From: Iván Sánchez Ortega <ivan at sanchezortega.es>
To: hot at openstreetmap.org
Cc: 
Sent: Monday, 5 November 2012, 15:19
Subject: Re: [HOT] Somalia import from 2011

On Domingo, 4 de noviembre de 2012 16:58:51 Kate Chapman escribió:
> I looked at the links you sent, I'm not sure why we would delete the
> data though. Yes it is imperfect, but I believe it was collected by
> GPS. The imagery at least suggests there are roads in those areas
> where you are pointing.
> 
> Schuyler who did the data conversion for the import might have some
> more insight.

Oh, I remember this. I think I've done about 30% of all the Somalia import, so 
I'll jump in:



> Lot of streets, but 98% i delete because it is not fitting.
> Only primary road are fitting.

98%??! Oh, c'mon. The imported data wasn't *that* bad. Yes, it has kinks; yes, 
it has tracks tagged as roads; yes, it has lots of that awful "highway=road" 
tag combination.

That doesn't mean that you should delete all that. If the problem is 
positional accuracy, then:

* Provide a data source with better accuracy. Now, you might think that Bing 
imagery is perfect and all that. Well, don't. Fitting aerial imagery to the 
ground truth is a non-trivial thing, and by my experience (and other OSMers' 
experiences), you cannot expect this from Bing.

* Provide GPS traces if you can. Lots. GPS traces in any given area will give 
you an idea of the *local* deformation, if any, of the aerial images. Emphasis 
on the "local". If a road is offset 10 meters, you cannot expect that offset to 
be the very same 20 kilometers away.

* Fit the data to that accurate data source. Don't delete data unless you're 
certain that the features aren't on the ground.


Please.


-- 
Iván Sánchez Ortega <ivan at sanchezortega.es> <ivan at geonerd.org>

Homer no function beer well without.
                                                              -- Homer Simpson

_______________________________________________
HOT mailing list
HOT at openstreetmap.org
http://lists.openstreetmap.org/listinfo/hot




More information about the HOT mailing list