Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[John Robert Peterson]

I hate to sound pesemeistic, and may have missed somthing, but...

Do we have any gaurentee that this inacuracy isn't some messed up copy
protection/obfucation/deliberate error/easter egg thing introduced by OS? I
know they shouldn't have, but they are a relitivly clumsy organisation, and
I'm sure there are people there that that don't like what's going on here,
there are all sorts of ways that this could have happened.

Also, are we sure that we are using a completly accurate reprojection back
form OS grid, with all of it's kinks and oddities?

JR

On 30 May 2010 23:32, James Hogan  wrote:

> On Sunday 30 May 2010 12:16:33 Tim François wrote:
> > By the way, which method are people using to re-project the
> VectorDistrict
> > data? I'm using the inbuilt datum in gdal - is anyone using the correct
> > *.prj file, and is there a difference?
>
> I'm using vectormap in Merkaartor from Git with the original prj files. I
> just
> tried in my area (york) with the modified prj file pointed to by Chris Hill
> and the difference is only around 30cm.
>
> The discrepancy between streetview tiles, vectormap and osm can be seen in
> this merkaartor screenshot:
> http://albanarts.com/~james/osm/pics/os_offsets_heworth.jpeg
>
> I don't think the reprojection of the original streetview tiles is the
> issue,
> as the buildlings generated by TimSC's script from the original tiles are
> only
> offset from the reprojected tiles by about 3m to the west here.
>
> Cheers
> James
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>
>
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[James Hogan]

On Sunday 30 May 2010 12:16:33 Tim François wrote:
> By the way, which method are people using to re-project the VectorDistrict
> data? I'm using the inbuilt datum in gdal - is anyone using the correct
> *.prj file, and is there a difference?

I'm using vectormap in Merkaartor from Git with the original prj files. I just 
tried in my area (york) with the modified prj file pointed to by Chris Hill 
and the difference is only around 30cm.

The discrepancy between streetview tiles, vectormap and osm can be seen in 
this merkaartor screenshot:
http://albanarts.com/~james/osm/pics/os_offsets_heworth.jpeg

I don't think the reprojection of the original streetview tiles is the issue, 
as the buildlings generated by TimSC's script from the original tiles are only 
offset from the reprojected tiles by about 3m to the west here.

Cheers
James


signature.asc
Description: This is a digitally signed message part.
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Robert Scott]

Funny this subject should come up.

Working with the OS Locator data, I'm finding the bounding boxes to be 
constantly about 50m to the east. In the south, they're also about 30m south. 
In scotland they're about 10m north.

I'm using the srid transform 27700 -> 4326.

What am I doing wrong?


robert.

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Steve Doerr]

From: "Chris Hill" 
Sent: Sunday, May 30, 2010 11:53 AM

> I believe the StreetView tiles are offset south(ish) by a few metres in
> East Yorkshire too. Reprojected shape files line up well with surveyed
> data. I have traced a few buildings from StreetView but I've stopped
> until I had worked out what was wrong. Now given other people's comments
> I do think there may be some discrepancy.
>
> Would a few carefully surveyed road junctions with many GPS traces to
> work from help to identify any discrepancy? Or is there a better way?

Are the OS StreetView tiles based on OS grid squares? If so, the 
co-ordinates of the four corners of any tile must be known with very great 
precision. Wouldn't give a basis for verification?

-- 
Steve 



___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Tim François]

So... I take that to mean that I should really be using the correct *.prj 
file, rather than gdal's inbuilt datum?

Tim

--- On Sun, 30/5/10, Jerry Clough - OSM  wrote:

From: Jerry Clough - OSM 
Subject: Re: [Talk-GB] Offsets between OS and OSM data (was Building with 
mapseg)
To: talk-gb@openstreetmap.org
Date: Sunday, 30 May, 2010, 12:30

Hi Tim,

Yes there is, at least with the version of GDAL I use. Chillly has written 
about this on his blog, and the changes needed (adding Helmert transformations 
- -sound fancy doesn't it) to the standard projection are noted in previous 
messages here in talk-gb.

I think the divergence is much greater on the E of the country: probably why 
Chilly and I worried most about it. Even with these the accuracy compared with 
the OSGB02 will be upto 5 metres out. See OS Coordinate Systems Guide.

Jerry

PS. StreetView and OSM seem to match up quite well for Nottingham. I've just 
rendered a set of tiles in OSGB36 of the same scale and boundaries as 
StreetView which
 at least removes some of the projection transformation artefacts.


From: Tim François 
To: Talk-GB@openstreetmap.org; Chris Hill 
Sent: Sun, 30 May, 2010 12:16:33
Subject: Re: [Talk-GB] Offsets between OS and OSM data (was Building with 
mapseg)

By the way, which method are people using to re-project the VectorDistrict
 data? I'm using the inbuilt datum in gdal - is anyone using the correct *.prj 
file, and is there a difference?

Tim

--- On Sun, 30/5/10, Chris Hill  wrote:

From: Chris Hill 
Subject: Re: [Talk-GB] Offsets between OS and OSM data (was Building with 
mapseg)
To: Talk-GB@openstreetmap.org
Date: Sunday, 30 May, 2010, 11:53

I believe the StreetView tiles are offset south(ish) by a few metres in 
East Yorkshire too. Reprojected shape files line up well with surveyed 
data. I have traced a few buildings from StreetView but I've stopped 
until I had worked out what was
 wrong. Now given other people's comments 
I do think there may be some discrepancy.

Would a few carefully surveyed road junctions with many GPS traces to 
work from help to identify any discrepancy? Or is there a better way?

Cheers, Chris

Kevin Peat wrote:
> I'm in Devon and I see the same thing although whether it is just the 
> SW I don't know. 
>
> The Streetview tiles (as I see them in JOSM) are all offset to the SE 
> by 5-10 metres. I've converted some woods in my area from the 
> VectorDistrict data using this process,
>
>     http://wiki.openstreetmap.org/wiki/Using_OS_Shapefiles
>
> and the converted data looks good to me compared to my previous 
> surveys but comes out different to the tiles, so I'm thinking that the 
> tiles are
 wrong.
>
> Kevin
>
>
>
>
> On 30 May 2010 09:08, Tim François  <mailto:sk1pp...@yahoo.co.uk>> wrote:
>
>     > On a side note, has anybody noticed a consistent tendency
>     > for existing
>     > independently surveyed roads to be offset northwards (by
>     > around 5-10
>     > metres) from the OS data (vectormap and streetview)? I've
>     > seen various
>     > cases of existing roads being edited to be consistent with
>     > OS data,
> 
    > but I'm not convinced this is a good idea since the problem
>     > seems to
>     > be consistent in one direction.
>
>     Glad I'm not the only one. Here in the SW I see the same offsets,
>     although I find the VectorDistrict data to be more like the GPS
>     surveyed data. This means that the StreetView tiles do not match
>     up with the VectorDistrict either: I've been importing some rivers
>     and reservoirs from the VectorDistrict data (namely the River Chew
>     and Chew Valley Lake) and I've found that the polygons seem to be
>     shifted compared to the equivalent positions in StreetView by
>     about 10 metres.
>
> 
    I guess this is an expected artifact of the reprojection methods?
>
>     Tim
>
>
>
>
>     ___
>     Talk-GB mailing list
>     talk...@openstreetmap.org <mailto:Talk-GB@openstreetmap.org>
>     http://lists.openstreetmap.org/listinfo/talk-gb
>
>
> 
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>   


___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb





   





  
-Inline Attachment Follows-

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb



  ___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Jerry Clough - OSM]

Hi Tim,

Yes there is, at least with the version of GDAL I use. Chillly has written 
about this on his blog, and the changes needed (adding Helmert transformations 
- -sound fancy doesn't it) to the standard projection are noted in previous 
messages here in talk-gb.

I think the divergence is much greater on the E of the country: probably why 
Chilly and I worried most about it. Even with these the accuracy compared with 
the OSGB02 will be upto 5 metres out. See OS Coordinate Systems Guide.

Jerry

PS. StreetView and OSM seem to match up quite well for Nottingham. I've just 
rendered a set of tiles in OSGB36 of the same scale and boundaries as 
StreetView which at least removes some of the projection transformation 
artefacts.






From: Tim François 
To: Talk-GB@openstreetmap.org; Chris Hill 
Sent: Sun, 30 May, 2010 12:16:33
Subject: Re: [Talk-GB] Offsets between OS and OSM data (was Building with 
mapseg)


By the way, which method are people using to re-project the VectorDistrict 
data? I'm using the inbuilt datum in gdal - is anyone using the correct *.prj 
file, and is there a difference?

Tim

--- On Sun, 30/5/10, Chris Hill  wrote:


>From: Chris Hill 
>Subject: Re: [Talk-GB] Offsets between OS and OSM data (was Building with 
>mapseg)
>To: Talk-GB@openstreetmap.org
>Date: Sunday, 30 May, 2010, 11:53
>
>
>I believe the StreetView tiles are offset south(ish) by a few metres in 
>East Yorkshire too. Reprojected shape files line up well with surveyed 
>data. I have traced a few buildings from StreetView but I've stopped 
>until I had worked out what was
> wrong. Now given other people's comments 
>I do think there may be some discrepancy.
>
>Would a few carefully surveyed road junctions with many GPS traces to 
>work from help to identify any discrepancy? Or is there a better way?
>
>Cheers, Chris
>
>Kevin Peat wrote:
>> I'm in Devon and I see the same thing although whether it is just the 
>> SW I don't know. 
>>
>> The Streetview tiles (as I see them in JOSM) are all offset to the SE 
>> by 5-10 metres. I've converted some woods in my area from the 
>> VectorDistrict data using this process,
>>
>> http://wiki.openstreetmap.org/wiki/Using_OS_Shapefiles
>>
>> and the converted data looks good to me compared to my previous 
>> surveys but comes out different to the tiles, so I'm thinking that the 
>> tiles are
> wrong.
>>
>> Kevin
>>
>>
>>
>>
>> On 30 May 2010 09:08, Tim François > <mailto:sk1pp...@yahoo.co.uk>> wrote:
>>
>> > On a side note, has anybody noticed a consistent tendency
>> > for existing
>> > independently surveyed roads to be offset northwards (by
>> > around 5-10
>> > metres) from the OS data (vectormap and streetview)? I've
>> > seen various
>> > cases of existing roads being edited to be consistent with
>> > OS data,
>> 
>> but I'm not convinced this is a good idea since the problem
>> > seems to
>> > be consistent in one direction.
>>
>> Glad I'm not the only one. Here in the SW I see the same offsets,
>> although I find the VectorDistrict data to be more like the GPS
>> surveyed data. This means that the StreetView tiles do not match
>> up with the VectorDistrict either: I've been importing some rivers
>> and reservoirs from the VectorDistrict data (namely the River Chew
>> and Chew Valley Lake) and I've found that the polygons seem to be
>> shifted compared to the equivalent positions in StreetView by
>> about 10 metres.
>>
>> 
>I guess this is an expected artifact of the reprojection methods?
>>
>> Tim
>>
>>
>>
>>
>> ___
>> Talk-GB mailing list
>> Talk-GB@openstreetmap.org <mailto:Talk-GB@openstreetmap.org>
>> http://lists.openstreetmap.org/listinfo/talk-gb
>>
>>
>> 
>>
>> ___
>> Talk-GB mailing list
>> Talk-GB@openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/talk-gb
>>   
>
>
>___
>Talk-GB mailing list
>Talk-GB@openstreetmap.org
>http://lists.openstreetmap.org/listinfo/talk-gb
> 



  ___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Tim François]

By the way, which method are people using to re-project the VectorDistrict 
data? I'm using the inbuilt datum in gdal - is anyone using the correct *.prj 
file, and is there a difference?

Tim

--- On Sun, 30/5/10, Chris Hill  wrote:

From: Chris Hill 
Subject: Re: [Talk-GB] Offsets between OS and OSM data (was Building with 
mapseg)
To: Talk-GB@openstreetmap.org
Date: Sunday, 30 May, 2010, 11:53

I believe the StreetView tiles are offset south(ish) by a few metres in 
East Yorkshire too. Reprojected shape files line up well with surveyed 
data. I have traced a few buildings from StreetView but I've stopped 
until I had worked out what was wrong. Now given other people's comments 
I do think there may be some discrepancy.

Would a few carefully surveyed road junctions with many GPS traces to 
work from help to identify any discrepancy? Or is there a better way?

Cheers, Chris

Kevin Peat wrote:
> I'm in Devon and I see the same thing although whether it is just the 
> SW I don't know. 
>
> The Streetview tiles (as I see them in JOSM) are all offset to the SE 
> by 5-10 metres. I've converted some woods in my area from the 
> VectorDistrict data using this process,
>
>     http://wiki.openstreetmap.org/wiki/Using_OS_Shapefiles
>
> and the converted data looks good to me compared to my previous 
> surveys but comes out different to the tiles, so I'm thinking that the 
> tiles are wrong.
>
> Kevin
>
>
>
>
> On 30 May 2010 09:08, Tim François  <mailto:sk1pp...@yahoo.co.uk>> wrote:
>
>     > On a side note, has anybody noticed a consistent tendency
>     > for existing
>     > independently surveyed roads to be offset northwards (by
>     > around 5-10
>     > metres) from the OS data (vectormap and streetview)? I've
>     > seen various
>     > cases of existing roads being edited to be consistent with
>     > OS data,
>     > but I'm not convinced this is a good idea since the problem
>     > seems to
>     > be consistent in one direction.
>
>     Glad I'm not the only one. Here in the SW I see the same offsets,
>     although I find the VectorDistrict data to be more like the GPS
>     surveyed data. This means that the StreetView tiles do not match
>     up with the VectorDistrict either: I've been importing some rivers
>     and reservoirs from the VectorDistrict data (namely the River Chew
>     and Chew Valley Lake) and I've found that the polygons seem to be
>     shifted compared to the equivalent positions in StreetView by
>     about 10 metres.
>
>     I guess this is an expected artifact of the reprojection methods?
>
>     Tim
>
>
>
>
>     ___
>     Talk-GB mailing list
>     talk...@openstreetmap.org <mailto:Talk-GB@openstreetmap.org>
>     http://lists.openstreetmap.org/listinfo/talk-gb
>
>
> 
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>   


___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb



  ___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Chris Hill]

I believe the StreetView tiles are offset south(ish) by a few metres in 
East Yorkshire too. Reprojected shape files line up well with surveyed 
data. I have traced a few buildings from StreetView but I've stopped 
until I had worked out what was wrong. Now given other people's comments 
I do think there may be some discrepancy.

Would a few carefully surveyed road junctions with many GPS traces to 
work from help to identify any discrepancy? Or is there a better way?

Cheers, Chris

Kevin Peat wrote:
> I'm in Devon and I see the same thing although whether it is just the 
> SW I don't know. 
>
> The Streetview tiles (as I see them in JOSM) are all offset to the SE 
> by 5-10 metres. I've converted some woods in my area from the 
> VectorDistrict data using this process,
>
> http://wiki.openstreetmap.org/wiki/Using_OS_Shapefiles
>
> and the converted data looks good to me compared to my previous 
> surveys but comes out different to the tiles, so I'm thinking that the 
> tiles are wrong.
>
> Kevin
>
>
>
>
> On 30 May 2010 09:08, Tim François  > wrote:
>
> > On a side note, has anybody noticed a consistent tendency
> > for existing
> > independently surveyed roads to be offset northwards (by
> > around 5-10
> > metres) from the OS data (vectormap and streetview)? I've
> > seen various
> > cases of existing roads being edited to be consistent with
> > OS data,
> > but I'm not convinced this is a good idea since the problem
> > seems to
> > be consistent in one direction.
>
> Glad I'm not the only one. Here in the SW I see the same offsets,
> although I find the VectorDistrict data to be more like the GPS
> surveyed data. This means that the StreetView tiles do not match
> up with the VectorDistrict either: I've been importing some rivers
> and reservoirs from the VectorDistrict data (namely the River Chew
> and Chew Valley Lake) and I've found that the polygons seem to be
> shifted compared to the equivalent positions in StreetView by
> about 10 metres.
>
> I guess this is an expected artifact of the reprojection methods?
>
> Tim
>
>
>
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org 
> http://lists.openstreetmap.org/listinfo/talk-gb
>
>
> 
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>   


___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Jason Cunningham]

I'm in London and using the method, linked to by Kevin, the VectorDistrict
looks very good. I also converted to kml to see how some some of the old
docks near sea level matched up in google earth and I was more than happy.
Overlaying Vector District and 'StreetView tiles' in Josm indicates the
Streetview tiles can be off in places

Jason

On 30 May 2010 10:09, Kevin Peat  wrote:

> I'm in Devon and I see the same thing although whether it is just the SW I
> don't know.
>
> The Streetview tiles (as I see them in JOSM) are all offset to the SE by
> 5-10 metres. I've converted some woods in my area from the VectorDistrict
> data using this process,
>
> http://wiki.openstreetmap.org/wiki/Using_OS_Shapefiles
>
> and the converted data looks good to me compared to my previous surveys but
> comes out different to the tiles, so I'm thinking that the tiles are wrong.
>
> Kevin
>
>
>
>
>
> On 30 May 2010 09:08, Tim François  wrote:
>
>> > On a side note, has anybody noticed a consistent tendency
>> > for existing
>> > independently surveyed roads to be offset northwards (by
>> > around 5-10
>> > metres) from the OS data (vectormap and streetview)? I've
>> > seen various
>> > cases of existing roads being edited to be consistent with
>> > OS data,
>> > but I'm not convinced this is a good idea since the problem
>> > seems to
>> > be consistent in one direction.
>>
>> Glad I'm not the only one. Here in the SW I see the same offsets, although
>> I find the VectorDistrict data to be more like the GPS surveyed data. This
>> means that the StreetView tiles do not match up with the VectorDistrict
>> either: I've been importing some rivers and reservoirs from the
>> VectorDistrict data (namely the River Chew and Chew Valley Lake) and I've
>> found that the polygons seem to be shifted compared to the equivalent
>> positions in StreetView by about 10 metres.
>>
>> I guess this is an expected artifact of the reprojection methods?
>>
>> Tim
>>
>>
>>
>>
>> ___
>> Talk-GB mailing list
>> Talk-GB@openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/talk-gb
>>
>
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>
>
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] Offsets between OS and OSM data (was Building with mapseg)

[Kevin Peat]

I'm in Devon and I see the same thing although whether it is just the SW I
don't know.

The Streetview tiles (as I see them in JOSM) are all offset to the SE by
5-10 metres. I've converted some woods in my area from the VectorDistrict
data using this process,

http://wiki.openstreetmap.org/wiki/Using_OS_Shapefiles

and the converted data looks good to me compared to my previous surveys but
comes out different to the tiles, so I'm thinking that the tiles are wrong.

Kevin




On 30 May 2010 09:08, Tim François  wrote:

> > On a side note, has anybody noticed a consistent tendency
> > for existing
> > independently surveyed roads to be offset northwards (by
> > around 5-10
> > metres) from the OS data (vectormap and streetview)? I've
> > seen various
> > cases of existing roads being edited to be consistent with
> > OS data,
> > but I'm not convinced this is a good idea since the problem
> > seems to
> > be consistent in one direction.
>
> Glad I'm not the only one. Here in the SW I see the same offsets, although
> I find the VectorDistrict data to be more like the GPS surveyed data. This
> means that the StreetView tiles do not match up with the VectorDistrict
> either: I've been importing some rivers and reservoirs from the
> VectorDistrict data (namely the River Chew and Chew Valley Lake) and I've
> found that the polygons seem to be shifted compared to the equivalent
> positions in StreetView by about 10 metres.
>
> I guess this is an expected artifact of the reprojection methods?
>
> Tim
>
>
>
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb