[OSM-talk] Help to import Rio de Janeiro city data to OSM - misaligned shapefiles

Iván Sánchez Ortega ivan at sanchezortega.es
Fri May 14 09:42:27 BST 2010


El 14/05/2010 5:30, Arlindo Pereira escribió:
> [...] using Iván's ogr2osm.py [1] and it worked out pretty well.

Oi!

I'm glad my software is useful. Remember that ogr2osm is beerware,
though :-)


> However, the tracks are misaligned with the tracks I already have
> collected with GPS

Yeah, it usually happens with mid-1900s datums. Here in Spain, data is
shifted 60 to 150 meters when reprojecting from "European Datum 1950"
(A.K.A. ED50). Your data is based on the 1969 South American ellipsoid,
so I'd expect a similar shift.


> Any ideas? I don't think that the government data is misaligned like
> that.

You're wrong: it is. It all comes down to which reference system you
use. Unfortunately, you'll need two years of geodetics classes in a
university to have a full understanding of the issue.


> I thought about creating a shell script to add/subtract the lat/lon
> numbers on all nodes, but maybe someone would come out with a magical
> conversion that would work out perfectly :)

That "magical conversion" is called a Proj.4 string. Instead of letting 
ogr2osm guess the projection, you can specify all of its parameters, 
including a manual shift on the x, y and z axis.

I do suggest you get in touch with some local expert in geodesy. They 
will probably have faced this problem before (SAD69 -> WGS84), and will 
tell you the best course of action (a fixed shift, a grid shift based on 
a nadgrids file, etc).


BTW, try "ogr2osm -e 29183" - or ask someone about reprojecting from 
EPSG:29183 to EPSG:4326. 29183 means SAD69 in UTM 23S, 4326 means WGS84 
in lat-lon.



A gente se ve,
-- 
Iván Sánchez Ortega <ivan at sanchezortega.es>




More information about the talk mailing list