Re: [Talk-ca] canvec data offset
Bonjour, I'll have a look to see if I can do something for it in the next release. I suspect that might be caused by data resolution. Lat and Lon are stored in decimal degrees with 7 digits precision (48.1234567). The 7th digit will provide a precision around 0.001 meter. The 6th digit a precision around 0.027 meter witch look like your 0.03 meter. Cheers Daniel -Original Message- From: michael bishop [mailto:cle...@nbnet.nb.ca] Sent: January 30, 2012 16:11 To: Talk-CA OpenStreetMap Subject: [Talk-ca] canvec data offset ive been working with canvec for a few days now, and ive noticed some of the data is offset by 0.03 meters its not matching up with nearby tiles for example 021O10 doesnt match up with 021O07 ive noticed the problem on other tiles aswell, but didnt think to write them down ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] canvec data offset
the south east corner of 021O10.0 is at 47.534 by -66.7500013 the north east corner of 021O07.1 (should be exact same node), is at 47.500 by -66.750 the exact offset differs from corner to corner, some are off more then others, some are off in a different direction On 01/31/2012 08:11 AM, Bégin, Daniel wrote: Bonjour, I'll have a look to see if I can do something for it in the next release. I suspect that might be caused by data resolution. Lat and Lon are stored in decimal degrees with 7 digits precision (48.1234567). The 7th digit will provide a precision around 0.001 meter. The 6th digit a precision around 0.027 meter witch look like your 0.03 meter. Cheers Daniel -Original Message- From: michael bishop [mailto:cle...@nbnet.nb.ca] Sent: January 30, 2012 16:11 To: Talk-CA OpenStreetMap Subject: [Talk-ca] canvec data offset ive been working with canvec for a few days now, and ive noticed some of the data is offset by 0.03 meters its not matching up with nearby tiles for example 021O10 doesnt match up with 021O07 ive noticed the problem on other tiles aswell, but didnt think to write them down ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] canvec data offset
I've seen some of these deviations as well during Canvec import. Because they are so small ( 1 meter), I just decided to glue the polygons together, so any slivers are gone. It is inconvenient though. Could it be related to that some sheets were originally still in NAD27, instead of NAD83 (which is approximately WGS84, as used by OSM)? Frank On 31-1-2012 15:06, michael bishop wrote: the south east corner of 021O10.0 is at 47.534 by -66.7500013 the north east corner of 021O07.1 (should be exact same node), is at 47.500 by -66.750 the exact offset differs from corner to corner, some are off more then others, some are off in a different direction On 01/31/2012 08:11 AM, Bégin, Daniel wrote: Bonjour, I'll have a look to see if I can do something for it in the next release. I suspect that might be caused by data resolution. Lat and Lon are stored in decimal degrees with 7 digits precision (48.1234567). The 7th digit will provide a precision around 0.001 meter. The 6th digit a precision around 0.027 meter witch look like your 0.03 meter. Cheers Daniel -Original Message- From: michael bishop [mailto:cle...@nbnet.nb.ca] Sent: January 30, 2012 16:11 To: Talk-CA OpenStreetMap Subject: [Talk-ca] canvec data offset ive been working with canvec for a few days now, and ive noticed some of the data is offset by 0.03 meters its not matching up with nearby tiles for example 021O10 doesnt match up with 021O07 ive noticed the problem on other tiles aswell, but didnt think to write them down ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] canvec data offset
Bonjour Frank, The problem is not related to a NAD27-NAD83 conversion. With Michael's example, I am now sure that the problem is related to the quadtree clipper. I must find a way to round quadtree's tiles coordinates to a proper value. Thanks to Michael Best regard, Daniel -Original Message- From: Frank Steggink [mailto:stegg...@steggink.org] Sent: January 31, 2012 11:56 To: talk-ca@openstreetmap.org Subject: Re: [Talk-ca] canvec data offset I've seen some of these deviations as well during Canvec import. Because they are so small ( 1 meter), I just decided to glue the polygons together, so any slivers are gone. It is inconvenient though. Could it be related to that some sheets were originally still in NAD27, instead of NAD83 (which is approximately WGS84, as used by OSM)? Frank On 31-1-2012 15:06, michael bishop wrote: the south east corner of 021O10.0 is at 47.534 by -66.7500013 the north east corner of 021O07.1 (should be exact same node), is at 47.500 by -66.750 the exact offset differs from corner to corner, some are off more then others, some are off in a different direction On 01/31/2012 08:11 AM, Bégin, Daniel wrote: Bonjour, I'll have a look to see if I can do something for it in the next release. I suspect that might be caused by data resolution. Lat and Lon are stored in decimal degrees with 7 digits precision (48.1234567). The 7th digit will provide a precision around 0.001 meter. The 6th digit a precision around 0.027 meter witch look like your 0.03 meter. Cheers Daniel -Original Message- From: michael bishop [mailto:cle...@nbnet.nb.ca] Sent: January 30, 2012 16:11 To: Talk-CA OpenStreetMap Subject: [Talk-ca] canvec data offset ive been working with canvec for a few days now, and ive noticed some of the data is offset by 0.03 meters its not matching up with nearby tiles for example 021O10 doesnt match up with 021O07 ive noticed the problem on other tiles aswell, but didnt think to write them down ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] canvec data offset
yeah, thats what i had to do for every node but part is the issue there, is that when i merge 2 nodes, it doesnt always go to the 'right' position' need to manualy select them in the right order to make it go in the right direction, it seems more like an error in the canvec data then an extra step i need to do on every tile almost looks like floating point rounding errors, but not sure why some tiles are ok and others arent On 01/30/2012 05:56 PM, Samuel Longiaru wrote: Have you tried just zooming back a bit in JOSM and merging them? It's part of my stitching process for each tile. I've never had a merge refused because of a 3 cm difference (but, of course I've never looked that closely to see how far apart they were actually). We're dealing with precision here... not accuracy at that level. Sam -Original Message- *From*: michael bishop cle...@nbnet.nb.ca mailto:michael%20bishop%20%3ccle...@nbnet.nb.ca%3e *To*: Samuel Longiaru longi...@shaw.ca mailto:samuel%20longiaru%20%3clongi...@shaw.ca%3e *Subject*: Re: [Talk-ca] canvec data offset *Date*: Mon, 30 Jan 2012 17:42:42 -0400 yeah, its not much, but its enough to screw up josm when trying to combine things between the 2 tiles have to fix each node on the border one by one, no real patern to which direction they are shifted even along a solid line (the edge of the tile), the nodes are going updown and zig-zagging across the edge On 01/30/2012 05:31 PM, Samuel Longiaru wrote: I can hardly put my finger to my nose to within .03 meters... and that's not even drunk. :) Sam L. Kamloops -Original Message- *From*: michael bishop cle...@nbnet.nb.ca mailto:michael%20bishop%20%3ccle...@nbnet.nb.ca%3e *To*: Talk-CA OpenStreetMap talk-ca@openstreetmap.org mailto:talk-ca%20openstreetmap%20%3ctalk...@openstreetmap.org%3e *Subject*: [Talk-ca] canvec data offset *Date*: Mon, 30 Jan 2012 17:11:27 -0400 ive been working with canvec for a few days now, and ive noticed some of the data is offset by 0.03 meters its not matching up with nearby tiles for example 021O10 doesnt match up with 021O07 ive noticed the problem on other tiles aswell, but didnt think to write them down ___ Talk-ca mailing list Talk-ca@openstreetmap.org mailto:Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] canvec data offset
But which is the rightposition? When one is given to be 3 cm's off the other, neither is more accurate, given the data collection technique. I think you are beyond the accuracy of the data themselves at that level. While merging and then combining a stream or a road that crosses a tile boundary, trying to decide which node is more accurate is one of those questions best asked only during periods of intensive navel contemplation. Take either one. The precision of the data exceed their accuracy I would guess by several orders of magnitude. Is the next node along the way mapped to 3 cm accuracy? Or the next, or the next? No. Here is an example where good enough is TRULY good enough. It's all the science can bear. Sam -Original Message- From: michael bishop cle...@nbnet.nb.ca To: Talk-CA OpenStreetMap talk-ca@openstreetmap.org Subject: Re: [Talk-ca] canvec data offset Date: Mon, 30 Jan 2012 18:01:04 -0400 yeah, thats what i had to do for every node but part is the issue there, is that when i merge 2 nodes, it doesnt always go to the 'right' position' need to manualy select them in the right order to make it go in the right direction, it seems more like an error in the canvec data then an extra step i need to do on every tile almost looks like floating point rounding errors, but not sure why some tiles are ok and others arent On 01/30/2012 05:56 PM, Samuel Longiaru wrote: Have you tried just zooming back a bit in JOSM and merging them? It's part of my stitching process for each tile. I've never had a merge refused because of a 3 cm difference (but, of course I've never looked that closely to see how far apart they were actually). We're dealing with precision here... not accuracy at that level. Sam -Original Message- From: michael bishop cle...@nbnet.nb.ca To: Samuel Longiaru longi...@shaw.ca Subject: Re: [Talk-ca] canvec data offset Date: Mon, 30 Jan 2012 17:42:42 -0400 yeah, its not much, but its enough to screw up josm when trying to combine things between the 2 tiles have to fix each node on the border one by one, no real patern to which direction they are shifted even along a solid line (the edge of the tile), the nodes are going updown and zig-zagging across the edge On 01/30/2012 05:31 PM, Samuel Longiaru wrote: I can hardly put my finger to my nose to within .03 meters... and that's not even drunk. :) Sam L. Kamloops -Original Message- From: michael bishop cle...@nbnet.nb.ca To: Talk-CA OpenStreetMap talk-ca@openstreetmap.org Subject: [Talk-ca] canvec data offset Date: Mon, 30 Jan 2012 17:11:27 -0400 ive been working with canvec for a few days now, and ive noticed some of the data is offset by 0.03 meters its not matching up with nearby tiles for example 021O10 doesnt match up with 021O07 ive noticed the problem on other tiles aswell, but didnt think to write them down ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] canvec data offset
i wrote a josm plugin that does all the math and marks the border of every canvec tile some of the tiles match my grid perfectly, and match eachother with no overlap or gap other tiles are offset slightly and dont fit the grid or eachother by checking the grid in my plugin, i can tell which node is in the 'correct' spot, because the canvec edges are always a certain fraction of a degree On 01/30/2012 06:36 PM, Samuel Longiaru wrote: But which is the rightposition? When one is given to be 3 cm's off the other, neither is more accurate, given the data collection technique. I think you are beyond the accuracy of the data themselves at that level. While merging and then combining a stream or a road that crosses a tile boundary, trying to decide which node is more accurate is one of those questions best asked only during periods of intensive navel contemplation. Take either one. The precision of the data exceed their accuracy I would guess by several orders of magnitude. Is the next node along the way mapped to 3 cm accuracy? Or the next, or the next? No. Here is an example where good enough is TRULY good enough. It's all the science can bear. Sam -Original Message- *From*: michael bishop cle...@nbnet.nb.ca mailto:michael%20bishop%20%3ccle...@nbnet.nb.ca%3e *To*: Talk-CA OpenStreetMap talk-ca@openstreetmap.org mailto:talk-ca%20openstreetmap%20%3ctalk...@openstreetmap.org%3e *Subject*: Re: [Talk-ca] canvec data offset *Date*: Mon, 30 Jan 2012 18:01:04 -0400 yeah, thats what i had to do for every node but part is the issue there, is that when i merge 2 nodes, it doesnt always go to the 'right' position' need to manualy select them in the right order to make it go in the right direction, it seems more like an error in the canvec data then an extra step i need to do on every tile almost looks like floating point rounding errors, but not sure why some tiles are ok and others arent On 01/30/2012 05:56 PM, Samuel Longiaru wrote: Have you tried just zooming back a bit in JOSM and merging them? It's part of my stitching process for each tile. I've never had a merge refused because of a 3 cm difference (but, of course I've never looked that closely to see how far apart they were actually). We're dealing with precision here... not accuracy at that level. Sam -Original Message- *From*: michael bishop cle...@nbnet.nb.ca mailto:michael%20bishop%20%3ccle...@nbnet.nb.ca%3e *To*: Samuel Longiaru longi...@shaw.ca mailto:samuel%20longiaru%20%3clongi...@shaw.ca%3e *Subject*: Re: [Talk-ca] canvec data offset *Date*: Mon, 30 Jan 2012 17:42:42 -0400 yeah, its not much, but its enough to screw up josm when trying to combine things between the 2 tiles have to fix each node on the border one by one, no real patern to which direction they are shifted even along a solid line (the edge of the tile), the nodes are going updown and zig-zagging across the edge On 01/30/2012 05:31 PM, Samuel Longiaru wrote: I can hardly put my finger to my nose to within .03 meters... and that's not even drunk. :) Sam L. Kamloops -Original Message- *From*: michael bishop cle...@nbnet.nb.ca mailto:michael%20bishop%20%3ccle...@nbnet.nb.ca%3e *To*: Talk-CA OpenStreetMap talk-ca@openstreetmap.org mailto:talk-ca%20openstreetmap%20%3ctalk...@openstreetmap.org%3e *Subject*: [Talk-ca] canvec data offset *Date*: Mon, 30 Jan 2012 17:11:27 -0400 ive been working with canvec for a few days now, and ive noticed some of the data is offset by 0.03 meters its not matching up with nearby tiles for example 021O10 doesnt match up with 021O07 ive noticed the problem on other tiles aswell, but didnt think to write them down ___ Talk-ca mailing list Talk-ca@openstreetmap.org mailto:Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org mailto:Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca