[Talk-gb-westmidlands] [Talk-transit] Naptan alignment

Wed Apr 1 09:42:41 BST 2009

Peter Miller [mailto:petermiller63 at googlemail.com] On Behalf Of Peter
Miller wrote:
>Sent: 01 April 2009 8:05 AM
>To: Andy Robinson (blackadder-lists)
>Cc: roger at slevin.plus.com; 'Brian Prangle'; Talk-gb-
>westmidlands at openstreetmap.org; talk-transit at openstreetmap.org; 'Thomas
>Wood'
>Subject: Re: [Talk-transit] [Talk-gb-westmidlands] Naptan alignment
>
>
>On 31 Mar 2009, at 23:43, Andy Robinson (blackadder-lists) wrote:
>
>
>	Roger Slevin [mailto:roger at slevin.plus.com] wrote:
>
>
>		Sent: 31 March 2009 11:20 PM
>
>
>		To: 'Andy Robinson (blackadder-lists)'; 'Brian Prangle';
Talk-
>gb-
>
>
>		westmidlands at openstreetmap.org;
talk-transit at openstreetmap.org;
>'Thomas
>
>
>		Wood'
>
>
>		Subject: RE: [Talk-transit] [Talk-gb-westmidlands] Naptan
>alignment
>
>
>
>		Andy
>
>
>
>		"Custom and Practice" stops - that is stops which are not
>physically marked
>
>
>		- are a common feature in many parts of the country ... less
so
>within
>
>
>		Metropolitan areas than in rural ones.  And many are not
>"signed in one
>
>
>		direction to represent both directions" - they are without
any
>sign.  These
>
>
>		unmarked stops are stop type CUS in NaPTAN.  I appreciate
that
>their
>
>
>		representation on a map might be conceptually challenging
...
>but they are
>
>
>		essential points as far as bus passengers are concerned!
>
>
>
>	I have no problem representing them on a map and indeed the whole
>concept is
>	fine. Our challenge is one of verification. We essentially only put
>data in
>	OSM that is physically there on the ground. However at least these
>stops
>	have the "CUS" tag so it should be possible to verify that busses
>stop at
>	the location and the evidence on the ground might then be the pile
of
>fag
>	ends in the gutter ;-)
>
>
>I would have thought that in the UK the 'verification' of a customary stop
>could initially be that the data is within NaPTAN and that it doesn't seem
>unreasonable (ie it is not on a one-way street). Cigarette ends and people
>standing around looking up the road might corroborate that information but
>I suggest that they are not necessary where NaPTAN has provided data.

I don't trust NaPTAN one jot ;-) It may be a great data set but until I've
seen evidence that it matches reasonably what's on the ground I won't trust
it any more than any other data source. At the moment I'm seeing a
reasonable level of difference between the positional data and the location
of the stop on the ground so we know at least this aspect needs work and
there may be other aspects. 

>
>We should note that this import is getting attention from professionals and
>that in the UK/ Transmodel because it is seen as 'The' essential link
>between the timetables and the physical world and they have seen and
>discussed a lot of unusual situations that need to  be accommodated in the
>past 10 years. I would suggest therefore that we see if we can accommodate
>their requirements.

I'm all ears, but my view is that the model needs to match the practice;
regardless of what someone previously thought was a good idea.

>
>Could I suggest that normally a bus-stop will be  a pole, a shelter, a
>customary stop or a combination of a pole/shelter combination where the
>elements are close together and where the bus stops near that node and the
>stop can be modelled using a single node at that point with the appropriate
>tags. 

Agreed, that’s our current focus. Having said that, a separate shelter away
from a posted bus stop sign may get a separate node to note it’s a
building=bus_shelter or something

However, In situations where the stopping position (as indicated by a
>lay-by, paint on the road or a section of raised curb) is different from
>the position of the shelter then the bus stop should be shown at the point
>where people enter the vehicle and that the shelter/pole should be mapped
>as a separate feature?

That’s probably not going to happen for the foreseeable future. When mapping
we locate and tag what we can see on the ground. If someone in the future
wants to be pedantic and sit and wait to see the location that buses stop as
a further verification that’s fine, but that’s not a priority for completing
basic level mapping. However I would think the best approach for this would
be to get the trace data from the bus operator or from riding the route to
evaluate the differences between stop signage position and setting down
position and see if further work is needed. I suspect that if it's within 5m
or so (and most will be) its really not worth the effort. 

>
>There are probably a couple of reasons that the professional community is
>interested in this detail.  Firstly with GPS tracked buses it is
>importation to know if the bus is crawling towards the bus stop through
>traffic and should be shown on the electronic sign as 'due' or if it is at
>the bus stop, or if it is crawling away from the stop through traffic and
>should not be displayed. 

The position the doors open at should be where the people stand to catch the
bus, otherwise it's an imperfect service. If we get the former right in the
long run then it should be more than good enough. In practice if a bus is
stuck in traffic it will normally open its doors once it's within 10 m of
the stop anyway simply because people are impatient.

Also... with GPS tracked buses the time that it
>'arrived' at the stop and 'left' the stop are both recorded for management
>purposes and bus companies can be fined £100K+ for failing to meet required
>timings so this is a matter of great importance to them! The other reason
>to be concerned about this is for blind/partially sighted.

Noted and so I'll modify my mapping behaviour a little for shelters and
ensure I map the open end of the shelter. Some older shelters in Birmingham
are canopy type so people tend to stand at the end closest to the arriving
bus. For partially closed shelters we should locate the stop at the position
of the open end.
If bus companies are potentially being fined £100k for poor timings then
they won't be too concerned about investing some cash to verify and improve
the data themselves if necessary. 

>
>I note that there is no aerial photography for Birmingham from yahoo. I
>also note from http://sautter.com/map  <http://sautter.com/map> that the
>alignment for roads in Bham on OSM is out by significant distances in some
>places when compared to TeleAtlas data on Google maps and that the
>TeleAtlas data aligns with the aerial photography on Google so it looks as
>though some roads in central Bham on OSM are slightly out for
>understandable reasons. This means that in some cases we will will need to
>move NaPTAN stops to the 'wrong' geolocation to get them to match correctly
>with the roads which is a shame. In other cases it is clear from looking at
>NaPTAN data on top of google aerial photography that the NaPTAN stops are
>already in the wrong position - sometimes the stops are offset back from
>the road and on other occasions they are incorrectly positioned along the
>road.
>
>So.... what is we were able to find someone to pay for rectified Digital
>Globe satellite photography for Bham to allow the detailed street geometry
>for the area in OSM to be corrected? The cost of this for the 700 sq km
>that makes up most of Bham would be able £6K. If that would be useful to
>OSM then possibly someone in the local transport authority or DfT could be
>persuaded to provide the necessary funds to purchase this, however the
>first question is if that would be appreciated and used by the local OSM
>community. I certainly found it useful in Ipswich to use yahoo aerial
>photography to adjust my physical survey and spot missing streets in my
>initial survey - and OSM and TeleAtlas are now much closed together for
>road alignment in Ipswich using sautter as a result. This seems to be the
>only practical and legal way to get the data right for NaPTAN and for OSM
>in Bham.

Any aerial imagery would be great provided it's properly rectified. The
Yahoo imagery in some towns is as much as 30m displaced so you can't be sure
that roads will be in the right position. My recent experience using a high
sensitivity GPS receiver at 1 sec recording interval produces very good
correlation. Just check out Walsall Wood, Rushall, Shelfield or the east,
south and west of Walsall outside the ringroad. Elsewhere the most likely
reason for poor correlation is a mixture of older GPS use and a sampling
rate on the GPS set to automatic or a long distance/time interval. This
results in loss of road shape quite quickly. Outside the Birmingham inner
ring road aerial imagery probably isn’t necessary if high sensitivity
receivers are used because we need to resurvey bus stops anyway and we'll
therefore get more traces. I suspect even now if we pull up traces for roads
that were mapped a long time ago then we can tweak the road alignment to
better fit the increased volume of trace data.

Personally I prefer not to use aerial imagery for the first survey pass,
beyond that though it's invaluable for doing all the other cool things like
adding buildings and landuse areas properly as well as tweaking all objects
that haven’t been precisely positioned, including bus stops if they are
discernable (shelters should be). Others will probably feel the opposite to
me in that it enables them to take out a printed map which can then be
annotated to gather all the additional data, it’s a matter of surveying
style and tools available. In terms of getting new participation in OSM,
aerial imagery lowers the barrier and allows a lot more people to add a lot
more data which we want to encourage, thus I'm very much for obtaining
imagery if we can.

Cheers

Andy

>
>By way of reassurance, yes you can use the DG photography to derive
>mapping, but you can't use the DG photography within public facing products
>itself for that price. We used this approach for the Gaza project and it
>worked fine. Here are some more details:
>https://www.swiftpage6.com/speasapage.aspx?X=2Y0QTEPLHQPGU9KC00Z5WD
>
>
>
>Regards,
>
>
>Peter
>

>
>
>
>
>
>
>
>
>
>
>	Cheers
>
>	Andy
>
>
>
>
>		Roger
>
>
>
>
>		-----Original Message-----
>
>
>		From: talk-transit-bounces at openstreetmap.org
>
>
>		[mailto:talk-transit-bounces at openstreetmap.org] On Behalf Of
>Andy Robinson
>
>
>		(blackadder-lists)
>
>
>		Sent: 31 March 2009 23:11
>
>
>		To: 'Brian Prangle'; Talk-gb-westmidlands at openstreetmap.org;
>
>
>		talk-transit at openstreetmap.org; 'Thomas Wood'
>
>
>		Subject: Re: [Talk-transit] [Talk-gb-westmidlands] Naptan
>alignment
>
>
>
>		Using http://sautter.com/map I did a comparison of the
>precisely positioned
>
>
>		stops I mapped this morning in the Aldridge area. Assuming
>Google has the
>
>
>		locations the same as the NaPTAN data then I would say about
>one in 5
>
>
>		NaPTAN
>
>
>		stops has something wrong with the location. Mostly a stop
is
>displaced
>
>
>		along the street. These errors are as much as 30m.
>
>
>
>		One interesting question relates to stops on the ground that
>exist only on
>
>
>		one side of the street but state they also pickup/drop on
the
>opposite side
>
>
>		of the street. The NapTan data contains two stops when on
the
>ground there
>
>
>		is only one physical (pole or shelter). In general the
NaPTAN
>data appears
>
>
>		to show the stops staggered on either side of the street
when
>in practice
>
>
>		passengers are going to wait opposite the bus stop
>sign/shelter. At the
>
>
>		moment I'm mapping these with one node and an opposite=yes
tag
>on them.
>
>
>		There is no way to map the stop on the opposite side as it
>doesn't
>
>
>		physically exist. So what to do about the NaPTAN data in
this
>case.
>
>
>
>		Cheers
>
>
>
>		Andy
>
>
>
>
>			-----Original Message-----
>
>
>			From: talk-gb-westmidlands-bounces at openstreetmap.org
>[mailto:talk-gb-
>
>
>			westmidlands-bounces at openstreetmap.org] On Behalf Of
>Brian Prangle
>
>
>			Sent: 31 March 2009 9:46 AM
>
>
>			To: Talk-gb-westmidlands at openstreetmap.org; talk-
>
>
>		transit at openstreetmap.org;
>
>
>			Thomas Wood
>
>
>			Subject: [Talk-gb-westmidlands] Naptan alignment
>
>
>
>			Thomas
>
>
>
>			I've also looked at Google maps and their alignment
is
>off too in exactly
>
>
>			the same way ours is in areas I know well and have
>surveyed, so I guess
>
>
>			it's down to the NaPTAN data. There are examples
where I
>know the bus
>
>
>		stops
>
>
>			are in a row along the street (Corporation Street
and
>Acocks Green
>
>
>		Village
>
>
>			for example) but NapTAN has one or two skewed from
the
>line by several
>
>
>			metres.  Currently I favour correcting the NapTAN
data
>to what we know on
>
>
>			the ground, but until a consensus emerges I'm laying
off
>the urge to
>
>
>			correct it.
>
>
>
>			Regards
>
>
>
>			Brian
>
>
>
>
>
>		_______________________________________________
>
>
>		Talk-transit mailing list
>
>
>		Talk-transit at openstreetmap.org
>
>
>		http://lists.openstreetmap.org/listinfo/talk-transit
>
>
>
>
>
>
>	_______________________________________________
>	Talk-transit mailing list
>	Talk-transit at openstreetmap.org
>	http://lists.openstreetmap.org/listinfo/talk-transit
>
>