OpenStreetMap Forum

The Free Wiki World Map

You are not logged in.

#1 2009-11-29 10:55:34

greencaps
Member
Registered: 2009-10-05
Posts: 423

Polish nort of north border

Hello, I was looking at Poland's north frontier and saw that it was not all cyrillic placenames above it.

Names like Zgoda, Rodowo and so look Polish to me.

Upon futher investigation osm data claimed them to be Polish:
31528747 Zgoda
31530949 Radowo
31531983 Kalinowiec
31532156 Gronowo
60238036 Мамоново
255302463 Gronowo

These are not the only ones. If you just move the map you will see much more.

Now what is wrong here? Should the border be more to the north?

Offline

#2 2009-11-29 11:47:38

rhn
Member
Registered: 2008-10-30
Posts: 214

Re: Polish nort of north border

I think two things are wrong here.
All Polish names are just a little bit North of the border. Ways seem to be off for about the same distance (judging by border disconnections). Either border is shifted South or everything else shifted North.
The other wrong thing is that the data source for Polish villages is prettyinaccurate, their positions are rounded to 0.2°, and this can't be easily fixed.

Offline

#3 2009-11-29 15:59:24

greencaps
Member
Registered: 2009-10-05
Posts: 423

Re: Polish nort of north border

rhn wrote:

.. wrong thing is that the data source for Polish villages is prettyinaccurate, their positions are rounded to 0.2°, and this can't be easily fixed.

Fixing the datasource may be difficult. But moving some villages in Potlash is easy.

If I knew where these places were situated in reality then I would move them to the right location by hand.

It is not that there are thousends of them. Mayby a twenty?

Offline

#4 2009-11-29 18:50:00

rhn
Member
Registered: 2008-10-30
Posts: 214

Re: Polish nort of north border

It looks like the border is wrong, if this: http://ump.waw.pl/ is right.
I'd leave it as it is, until someone imports better border data.

Offline

#5 2009-11-30 10:55:15

balrog-kun
Member
From: Warsaw, Poland
Registered: 2008-08-10
Posts: 1,365
Website

Re: Polish nort of north border

greencaps wrote:

It is not that there are thousends of them. Mayby a twenty?

Altogether there are about 40000 villages (out of 60000) whose coordinates still need to be adjusted.

Some of them (perhaps some 1/2) can be fixed using coordinates from UMP but the remaining bunch are not in UMP.  I have a script to compare village names + coordinates + other info between UMP and OSM and adjust coordinates, should I run it on all of poland.osm?  So far I only ran it on some regions.

Offline

#6 2009-11-30 15:32:21

Zbigniew Czernik
Member
Registered: 2009-09-12
Posts: 1,308
Website

Re: Polish nort of north border

balrog-kun wrote:

Altogether there are about 40000 villages (out of 60000) whose coordinates still need to be adjusted.

Some of them (perhaps some 1/2) can be fixed using coordinates from UMP but the remaining bunch are not in UMP.  I have a script to compare village names + coordinates + other info between UMP and OSM and adjust coordinates, should I run it on all of poland.osm?  So far I only ran it on some regions.

In my opinion - no...  or only on untouched regions.

Offline

#7 2009-11-30 18:40:27

rhn
Member
Registered: 2008-10-30
Posts: 214

Re: Polish nort of north border

Zbigniew Czernik wrote:

In my opinion - no...  or only on untouched regions.

I second that

Offline

#8 2009-11-30 19:19:05

greencaps
Member
Registered: 2009-10-05
Posts: 423

Re: Polish nort of north border

rhn wrote:

I'd leave it as it is, until someone imports better border data.

balrog-kun wrote:

.. I have a script to compare village names + coordinates + other info between UMP and OSM and adjust coordinates, should I run it on all of poland.osm? ....

For me it looked as if you had found your man.

I do not understand why one does not second a correction of about 10 km for about 20.000 villages. What is the reason for not wanting that? Or is the data in UMP not correct?

Edit: do I have to conclude that not only so many villages lay a 10 km off position but also all roads leading to them ???

Last edited by greencaps (2009-11-30 19:21:53)

Offline

#9 2009-11-30 21:43:04

balrog-kun
Member
From: Warsaw, Poland
Registered: 2008-08-10
Posts: 1,365
Website

Re: Polish nort of north border

greencaps wrote:

Edit: do I have to conclude that not only so many villages lay a 10 km off position but also all roads leading to them ???

Nope, the villages whose coordinates are rounded were imported from either GNS or Geonames a long time ago and afaik nothing else was imported so it's only the vilages smile

Offline

#10 2009-12-02 11:14:27

rhn
Member
Registered: 2008-10-30
Posts: 214

Re: Polish nort of north border

Well, we surely can adjust all the villages to better fit UMP data, but *some* villages have better positions in OSM. That's why it only makes sense where the villages are important for some reason.

In this case, the *border* seems to be off, and the villages are good enough. We could correct them, but it would not change much, they'd still be North of the border. That's why I think we should get and verify some border data, and then import it. Any volunteers?

Offline

#11 2009-12-02 12:09:47

balrog-kun
Member
From: Warsaw, Poland
Registered: 2008-08-10
Posts: 1,365
Website

Re: Polish nort of north border

rhn wrote:

Well, we surely can adjust all the villages to better fit UMP data, but *some* villages have better positions in OSM.

The script doesn't touch villages that have ever been modified, only those that haven't been moved since the 2007 import by user:devrise so their coordinates are rounded.

rhn wrote:

That's why I think we should get and verify some border data, and then import it. Any volunteers?

Only problem is where do we take it from :)

Maybe Russia has some official public data about their borders and we could load Kaliningrad Oblast southern border from there, guess I can ask on the russian list.

Offline

Board footer

Powered by FluxBB