Re: [Talk-GB] bing image alignment
On Fri, Sep 13, 2013 at 08:09:49PM +0100, Andy Mabbett wrote: On 13 September 2013 19:59, ael law_ence@ntlworld.com wrote: No one has mentioned the OS gps (passive) stations: for example http://www.openstreetmap.org/browse/node/472420260 It's rather ironic that that node carries the note: Position from gps waypoint. Not from OS data which is more accurate. That was me being lazy :-) ael ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On Sat, Sep 14, 2013 at 02:24:31PM +0100, OpenStreetmap HADW wrote: On 13 September 2013 19:59, ael law_ence@ntlworld.com wrote: No one has mentioned the OS gps (passive) stations: for example http://www.openstreetmap.org/browse/node/472420260 http://www.ordnancesurvey.co.uk/business-and-government/help-and-support/navigation-technology/os-net/surveying.html Can we not use them at least for some reference points? Interesting. Unfortunately it looks as though OS Net slipped past their commercial people as they use the term freely available on their web site without giving a precise legal definition of what that means. Given the intended use (all high accuracy surveys in the UK) there seems to be an implication that they aren't claiming a database copyright, but OSM will probably need a clearer legal statement. OSTN02 seems to have the same licensing uncertainty. It is a few years since I looked at any of this, but it had not occurred to me that any copyright issue could arise. They are essentially modern trig points. There is a mark on the ground, and their website publishes the coordinates. Using these coordinates to position a node on osm would constitute republishing? I thought very small extracts of copyrighted material were permitted in any case. Depending on exactly which ground feature represents the station, That one is a small stud at ground level on a small concrete block all surrounded by a rectangular metal fence maybe 30cm high, roughly 1m by 1/2m. I must dig out a photograph. Are the ETRS89 coordinates given on the monument itself, as they would As far as I recall, there was no indication whatsoever of what it was nor any text at all. It's a pity they aren't all clearly visible on the aerial view, my local one, which appears to be within centimetres of Bing, is only I would be very surprised if the one at Buckland could be seen in aerial images. I must check. It is quite easy to miss it on the ground unless you know what you are looking for and where to look. The tall grass was enough to hide it. The biggest problem seems to be that there are only a few on public land and widely scattered, so of limited use. ael ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On Sat, Sep 14, 2013 at 02:24:31PM +0100, OpenStreetmap HADW wrote: On 13 September 2013 19:59, ael law_ence@ntlworld.com wrote: No one has mentioned the OS gps (passive) stations: for example http://www.openstreetmap.org/browse/node/472420260 http://www.ordnancesurvey.co.uk/business-and-government/help-and-support/navigation-technology/os-net/surveying.html It's a pity they aren't all clearly visible on the aerial view, my I have just checked, and I think the one at Buckland may be visible, but I need to check with the photographs that I took 4 years ago. in the current Bing images. I think that you can make out a small rectangular enclosure behind the crash barrier, and if I am right that is the metal fence that I mentioned. The stud of radius maybe 1 or 2cm is in the centre of the enclosure. All this from memory: need to find the photographs. ael ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 15 September 2013 11:27, ael law_ence@ntlworld.com wrote: It is a few years since I looked at any of this, but it had not occurred to me that any copyright issue could arise. They are essentially modern trig points. There is a mark on the ground, and their website publishes the coordinates. Using these coordinates to position a node on osm would constitute republishing? I thought very small extracts of copyrighted material were permitted in any case. It is the standard copyright issue for maps. There is a relatively new sort of copyright called database rights. Normally for copyright you need to have some creative input. Things like addresses and telephone numbers do not have any creative input. However database copyrights mean that the PAF (postcode address file) is copyright, as is your local telephone directory. With a map, the printed form may be covered by a copyright on the typographical arrangement, but the individual coordinates of the corners of a building are matters of fact and can't, individually be copyrighted. Nonetheless, once you start plotting many buildings, the work involved in compiling the information is recognized by a database copyright. Without database copyrights, there would be on copyright blocks on converting printed format maps to vector format. If you extract a single address from the PAF, to send letter, or a single number, from the phone book, to make a call, you are extracting the fact. If you create a directory, you are copying the database. The former is unrestricted (except by any contract for confidentiality). The latter is a copyright infringement. The OS owns the database copyright on its list of passive stations. Using the location of one of them for an isolated survey is just using a fact. Adding large numbers of them to OSM is a copyright infringement unless there is a licence that permits it. Doing so piecemeal still creates an infringement (if that were not the case, almost all use of commercial maps would be fair game, for a crowd sourced map). There are hints that the intent was an intent to licence with at most the equivalent of CC BY, but I've not seen anything that makes that explicit. On the other hand some of the references to OSTN02, which is actually a larger database, explicitly assert the database copyright, but also seem to release under a BSD licence, which, at most, would require some slight tweaking of copyright attributions. Please do not do anything based on this interpretation without verifying it yourself. (The copyright on OSM itself is essentially a database copyright.) Depending on exactly which ground feature represents the station, That one is a small stud at ground level on a small concrete block all surrounded by a rectangular metal fence maybe 30cm high, roughly 1m by 1/2m. I must dig out a photograph. Having seen OS' photograph, I'm fairly sure this is the feature at 5.2m from the GPS plot, when viewed on the local Bing datum. ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On Sat, Sep 14, 2013 at 02:24:31PM +0100, OpenStreetmap HADW wrote: On 13 September 2013 19:59, ael law_ence@ntlworld.com wrote: No one has mentioned the OS gps (passive) stations: for example http://www.openstreetmap.org/browse/node/472420260 http://www.ordnancesurvey.co.uk/business-and-government/help-and-support/navigation-technology/os-net/surveying.html Depending on exactly which ground feature represents the station, BIng Photographs of the Buckland station are here: http://homepage.ntlworld.com/lawrence_a_e/osm/ You can see part of the metal fence = railings in the 2nd picture. It seems to be a granite block or something similar, rather than concrete. I don't seem to have taken a wider view, so I am not absolutely sure that this is directly behind the crash barrier in the Bing image, but I think it is. ael ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 13 September 2013 19:59, ael law_ence@ntlworld.com wrote: No one has mentioned the OS gps (passive) stations: for example http://www.openstreetmap.org/browse/node/472420260 http://www.ordnancesurvey.co.uk/business-and-government/help-and-support/navigation-technology/os-net/surveying.html Can we not use them at least for some reference points? Interesting. Unfortunately it looks as though OS Net slipped past their commercial people as they use the term freely available on their web site without giving a precise legal definition of what that means. Given the intended use (all high accuracy surveys in the UK) there seems to be an implication that they aren't claiming a database copyright, but OSM will probably need a clearer legal statement. OSTN02 seems to have the same licensing uncertainty. I would note that even if a relatively restrictive licence were to apply, OS allow checking of accuracy, as long as you don't use the measured error to correct the inaccurate data. Depending on exactly which ground feature represents the station, BIng datum), Bing and the consumer grade GPS survey differ by between 5.2 and 6.2m. Bing and StreetView agree quite well at that point, but diverge towards the West. The BIng error in the typical range, at this point. Are the ETRS89 coordinates given on the monument itself, as they would be individual statements of fact, much like an individual address with postcode? (OS Net allows for movement, so the highest accuracy values will not be fixed.) Without survey quality GNSS receivers, these points are probably mainly of use for calibrating imagery. It's a pity they aren't all clearly visible on the aerial view, my local one, which appears to be within centimetres of Bing, is only identifiable because OS have measured the distance from the local trees. When the rain stops, I'll have a look at what is on the ground, and try for a long average GPS reading. I'll also try and get on the ground measurements from the trees in case the station coordinates are considered fair game, but the sketch map details aren't (looks like a surveyor's tape is on the shopping list - I guess class III, at about 0.05% is good enough, compared with GPS). I note that a lot of them are on private land and require permission from the landowner to gain physical access. PS. The OSTN02 conversion tool, which presumably contains the table, has a BSD Licence, according to http://www.ordnancesurvey.co.uk/business-and-government/help-and-support/navigation-technology/os-net/grid-inquest.html. If that is correct, I can see no reason why the os.openstreetmap tiles should not be rectified based on OSTN02. The data also seems to be available in lat/long to lat/long forrmat, also with a BSD licence. ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 10 September 2013 10:09, o...@k3v.eu wrote: This has been discussed on the list before. Bing image alignment can be quite poor One unfortunate consequence of this is that there are areas of the map where the majority of features are out of position by 15 or more feet, because people have added large numbers of buildings, from Bing, whilst treating the Bing locations as gospel. A couple of days ago, I walked a footpath, which turned out, along with the preceding private road, to be a PROW on foot. As I wanted to detail map it to show steps, I first calibrated Bing against OS StreetView, and moved both the road and path accordingly. The result is that it now goes right though the last building to actually be mapped on the road where the path ends. Looking at all the, un-sourced, (another gripe is the amount of material going in with no source and often no comment, but often quite extensive) buildings in the area, they seem to align perfectly with uncorrected Bing images. It is easy to perpetuate such problems by using the existing buildings to calibrate further additions. I'm reluctant to correct such bulk errors unless I can find somewhere to safely sit a GPS for several hours, with a clear view of the sky and good landmarks on aerial imagery photographs, to get absolute proof of the error. (I don't plan to hire a DGPS receiver!). Also, without an explicit source, everything needs carefully checking to make sure that it is consistent with the same flawed source. Often the main roads are more accurate in these areas, because they come from older sources, although I sometimes wonder what those were, too. ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
Which is the higher priority, consistency or accuracy? Is it better to have an internally consistent map, where everything is topologically correct but possibly a little displaced by a uniform vector, or is it better to have some of the objects positioned with high accuracy, despite the apparent conflicts with the rest? If we are going to get fussy about the accuracy of OSM objects, things aren't going to get mapped. My vote would go with a consistency based approach; at least all the objects will be wrong by the same amount, which might even be machine-correctable. But I will probably get shot down ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
Why do you suppose OS Streetview is correct? I find that compared to multiple GPS tracks it is not always well aligned and more recent Bing imagery is often better. OpenStreetmap HADW osmh...@gmail.com wrote: On 10 September 2013 10:09, o...@k3v.eu wrote: This has been discussed on the list before. Bing image alignment can be quite poor One unfortunate consequence of this is that there are areas of the map where the majority of features are out of position by 15 or more feet, because people have added large numbers of buildings, from Bing, whilst treating the Bing locations as gospel. A couple of days ago, I walked a footpath, which turned out, along with the preceding private road, to be a PROW on foot. As I wanted to detail map it to show steps, I first calibrated Bing against OS StreetView, and moved both the road and path accordingly. The result is that it now goes right though the last building to actually be mapped on the road where the path ends. Looking at all the, un-sourced, (another gripe is the amount of material going in with no source and often no comment, but often quite extensive) buildings in the area, they seem to align perfectly with uncorrected Bing images. --- cheers, Chris osm user, chillly ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 13 September 2013 12:31, Colin Smale colin.sm...@xs4all.nl wrote: Which is the higher priority, consistency or accuracy? Is it better to have an internally consistent map, where everything is topologically correct but possibly a little displaced by a uniform vector, or is it better to have some of the objects positioned with high accuracy, despite the apparent conflicts with the rest? The case in question is not internally consistent! The main roads are more correctly placed than the buildings. This doesn't stand out like a sore thumb because the roads and pavements are wide enough to take up the slack. However, a 1m wide footpath, sandwiched between two buildings, and displaced by about 6m does stand out. The other risk is of a gradual, but not as gradual as continental drift, creep as features get added relative to misplaced features placed relative to misplaced features. ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
OpenStreetmap HADW wrote: A couple of days ago, I walked a footpath, which turned out, along with the preceding private road, to be a PROW on foot. As I wanted to detail map it to show steps, I first calibrated Bing against OS StreetView When using any OS data it's important to be certain of the method used for the datum conversion. If it's not using the OSTN02 table-driven conversion you should expect to see offsets. The OS documentation says that the less-accurate method is OK for up to +/- 5m, but my experience is that the offset can be greater than that, e.g. up to about 20m. If you ever see offsets of this magnitude in anything that has touched OS grid references or OSGB36 datum, be suspicious! (As an example, http://hills-database.co.uk/ *deliberately* uses the less-accurate conversion because they claim that GPS devices will typically use the less-accurate conversion back to WGS84 when you enter a grid reference. As a consequence, if you display their summit positions on a not-reprojected OS map, they will be in the wrong places.) Cheers, Phil. ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 13 September 2013 12:55, Chris Hill o...@raggedred.net wrote: Why do you suppose OS Streetview is correct? I find that compared to multiple GPS tracks it is not always well aligned and more recent Bing imagery is often better. I did consider that possibility, but I did a search for that and it came up suggesting StreetView was mostly good to about 1m (although I can't re-find that reference). On the other hand, that is a reason for requiring more evidence before correcting things which assume zero offset for Bing. As noted elsewhere on this thread, Bing clearly shows parallax errors, and it looks like it hasn't been corrected for height variations, at least not on a local scale. I don't think I would trust commercial GPS much below 5m unless it was averaged over at least an hour, with a clear view of the sky (reflections off builidngs could systematically distort the position solution - see http://wiki.openstreetmap.org/wiki/Accuracy#Systematic_errors). Where I've looked into calibrating against GPS I've tended to find that there are good places that are accurately locatable on Bing where you can sit with GPS for a long time without looking suspicious, and which are near to points of interest and have a clear view of the sky. On the other hand, pre-OpenData specifications for StreetView suggest worst case errors of 4.1m www.centremapslive.co.uk/files/street_view_userguide.pdf, so, maybe I'm still going to have to do long averaged GPS calibrations. That still doesn't mean that unsourced BIng tracings, with zero offsets are good things. Fortunately, OSM hasn't been around enough for continental drift to become a severe problem. Even the error since 1984 should be less than about one foot. ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
Cm-level GPS accuracy is coming within our grasp... My attention was recently drawn to this: http://www.kickstarter.com/projects/swiftnav/piksi-the-rtk-gps-receiver [2] On 2013-09-13 15:06, OpenStreetmap HADW wrote: I don't think I would trust commercial GPS much below 5m unless it was averaged over at least an hour, with a clear view of the sky (reflections off builidngs could systematically distort the position solution - see http://wiki.openstreetmap.org/wiki/Accuracy#Systematic_errors [1]). Where I've looked into calibrating against GPS I've tended to find that there are good places that are accurately locatable on Bing where you can sit with GPS for a long time without looking suspicious, and which are near to points of interest and have a clear view of the sky. Links: -- [1] http://wiki.openstreetmap.org/wiki/Accuracy#Systematic_errors [2] http://www.kickstarter.com/projects/swiftnav/piksi-the-rtk-gps-receiver ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 13/09/13 14:28, OpenStreetmap HADW wrote: How does one find that out for the tiles served by os.openstreetmap.org, and if they are not using the high accuracy conversions, why not? (OS' own online viewer also uses WGS84 coordinates to label their tiles - suggesting that they should be using the high quality WGS-84 calibration.) They are produced using mapserv from a gdal vrt with the source projection defined by: PROJCS[OSGB 1936 / British National Grid, GEOGCS[OSGB 1936, DATUM[OSGB_1936, SPHEROID[Airy 1830,6377563.396,299.324964644, AUTHORITY[EPSG,7001]], TOWGS84[375,-111,431,0,0,0,0], AUTHORITY[EPSG,6277]], PRIMEM[Greenwich,0], UNIT[degree,0.0174532925199433], AUTHORITY[EPSG,4277]], PROJECTION[Transverse_Mercator], PARAMETER[latitude_of_origin,49], PARAMETER[central_meridian,-2], PARAMETER[scale_factor,0.9996012717], PARAMETER[false_easting,40], PARAMETER[false_northing,-10], UNIT[metre,1, AUTHORITY[EPSG,9001]], AUTHORITY[EPSG,27700]] Tom -- Tom Hughes (t...@compton.nu) http://compton.nu/ ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
AFAIK all OS StreetView tiles on OSM are projected in Spherical Mercator and this reprojection will certainly have been done using proj4, i.e, it's algorithmic not table-driven. OSSV scale is 1 pixel / metre, so accuracy is less than that. So sources of error are: - Feature generalisation in OSSV, noticeable on buildings roads - Re-projection of OSSV tiles using proj4 using OSGB36, errors of +/- 5m - Known GPS signal errors (LDOP) - Unknown systematic GPS errors - Fudge factor in aligning these two sources - Human factors in interpreting alignments Taken together this suggests 6-10 metres is entirely in the range that we might expect that keeping topological relationships is more important. Given an average road is about 7-8 metres across we are less likely to notice this with roads anyway. Obviously an aerial imagery source introduces a number of other error sources. If one looks at GPS traces for the same footpath walked again and again, their spread is quite considerable, perhaps as much as 20 m, although an average would probably be close to the actual 2 m path. So just like any other survey organisation, ultimately we will need our own set of convenient reference points. Jerry On 13 September 2013 14:37, OpenStreetmap HADW osmh...@gmail.com wrote: On 13 September 2013 14:22, Colin Smale colin.sm...@xs4all.nl wrote: Cm-level GPS accuracy is coming within our grasp... My attention was recently drawn to this: http://www.kickstarter.com/projects/swiftnav/piksi-the-rtk-gps-receiver You need to operate in differential mode to get the full accuracy, which means you still need an accurately surveyed datum. I don't see that it will help with reflections. I also wonder if it will prompt the re-introduction of selective availability, as it gives some of the advantages of the higher chipping rates on the, encrypted, find mode GPS signal, and probably doesn't have the velocity and altitude caps normally applied to commercial receivers. I believe the military have the advantage that they can use the second frequency to help do large area ionosphere corrections. ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 13 September 2013 14:46, Tom Hughes t...@compton.nu wrote: On 13/09/13 14:28, OpenStreetmap HADW wrote: How does one find that out for the tiles served by os.openstreetmap.org, and if they are not using the high accuracy conversions, why not? (OS' own online viewer also uses WGS84 coordinates to label their tiles - suggesting that they should be using the high quality WGS-84 calibration.) They are produced using mapserv from a gdal vrt with the source projection Does mapserv use OSTN02? defined by: PROJCS[OSGB 1936 / British National Grid, ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On Tue, 10 Sep 2013 09:29:24 +0100, Lester Caine les...@lsces.co.uk wrote: One reason for this is parallax error, because the images aren't taken square on to the ground (that may be because the camera is taking in quite a large area. You can see this with building, you can end up with a metre or more difference depending on whether you use the top or bottom of the building. It also presumably means that alignment changes with the height of the land. Obviously this pass is well over from the last one, which was pretty well aligned. There is quite a steep slope on the area I was working on last night and I can align things to the bottom or the top of the slope ... -1.68; -5.13 at the top -6.92; -8.13 down the slop :) This has been discussed on the list before. Bing image alignment can be quite poor where there are steep elevation changes. I see this a lot around the Devon coast where a lot of the settlements are on steep slopes. Google seem to handle this a lot better. Whether they spend more capturing the imagery or have some better way of post processing it who knows? Kevin ___ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] bing image alignment
On 9 September 2013 20:05, Lester Caine les...@lsces.co.uk wrote: I'm currently playing in an area where the highest resolution imagery is still an older view, while as I zoom out we step to newer imagery which is some distance off from the map tracks. I'm fairly happy with the map as I have had some older gps tracks which it follows, and I'll run over in the morning and gather a new track as a cross reference, but are people in general finding that these new images are out of alignment with what is currently mapped? Can I assume that they need realigning before using them? I find that current BIng is usually mis-allgned by two or three metres relative to OS StreetView, which I believe is good to better than 1 metre. One reason for this is parallax error, because the images aren't taken square on to the ground (that may be because the camera is taking in quite a large area. You can see this with building, you can end up with a metre or more difference depending on whether you use the top or bottom of the building. It also presumably means that alignment changes with the height of the land. On the the other hand, individual GPS points usually have larger error than this. With commercial grade GPS, you probably need several hours averaging to get down to a metre accuracy. ___ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
