Andy_J_Robinson at blueyonder.co.uk
Wed Aug 16 10:15:18 BST 2006
That's not surprising as the Landsat data position is not that accurate.
There is a shift in the Landsat. In my Birmingham area I normally have to
slide items created off the Landsat imagery (railway lines and rivers etc)
by about 50m to the south west.
Andy_J_Robinson at blueyonder.co.uk
>From: talk-bounces at openstreetmap.org [mailto:talk-
>bounces at openstreetmap.org] On Behalf Of Steve Chilton
>Sent: 16 August 2006 09:52
>To: talk at openstreetmap.org
>Subject: RE: [OSM-talk] Coastlines
>Further investigation leads me to think there might be some mismatch
>problems with this coastline data.
>If you zoom in in OSM and compare the alignment to the satelite image it
>appears to be misaligned.
>Following screenshot suggests that it is off slightly in a southerly
>direction. If you look at the truncated yellow roads they seem to align
>well with underlying image. One would really need higher-res satelite
>imagery to reach a more scientific conclusion.
>It is even more evident if you overlay the data over Google map, where
>it appears to be all shifted to south-east.
>What is missing is any inland detail to verify against. This does in
>fact point up the poor quality of this particular Google map data. See
>for example how the map shows the large offshore island as being part of
>the mainland, encompassing the smaller island on the way!
>What I suggest would be useful would be to convert the dataset for the
>Isle of White, incorporate within OSM, and allow cross checking there
>against GPS data, OSM satelite data and Google map data. Is that
>posible? What does anyone think?
>I still think this might be a useful dataset, but we need to know what
>we are getting if we accept it on a larger scale as a coastline dataset.
>Steve Chilton, Learning Support Fellow
>Learning and Technical Support Unit Manager
>School of Health and Social Sciences
>phone/fax: 020 8411 5355
>email: steve8 at mdx.ac.uk
>Chair of the Society of Cartographers:
>2006 SoC Summer School at Keele Uni:
>From: dankarran at gmail.com [mailto:dankarran at gmail.com] On Behalf Of Dan
>Sent: Tuesday, August 15, 2006 6:41 PM
>To: Steve Chilton
>Cc: talk at openstreetmap.org
>Subject: Re: [OSM-talk] Coastlines
>On 8/15/06, Steve Chilton <S.L.Chilton at mdx.ac.uk> wrote:
>> Data quality looks quite reasonable on this sample. I have made SVG
>> (saved as PNG) of the mainland bit of the example below.
>> There are two roads that go in the sea (1 in Port Erin and 1 in Port
>> Mary) but local gps work and editing could cure that. On balance I
>> say "better in than out" (ie if upload has worked for whole of island
>> then repeat in stages for other areas). Is projection an issue over
>> larger areas?
>The road in Port Erin is based not on GPS tracks but on the Landsat, and
>the one in Port St Mary is the harbour I think. There are a couple of
>other close calls from real GPS data, but they coastline around those
>areas is easily fixed.
>> NB: renderer (osmarender2) doesn't recognise all of linework as some
>> is still in class= rather than highway= format. Maybe worthwhile to
>> also check in a more (data) populated area. Has Douglas been mapped in
>> detail? Maybe I'll have a look at that when I have time.
>There are some bits around Douglas, but not as much as Port Erin/Port St
>Mary yet. I should go through and recategorise PSM as Rob put a lot of
>work into that and it'd be a shame for it not to show up when rendered.
>dan at karran.net
>talk mailing list
>talk at openstreetmap.org
More information about the talk