Imagery variations/misalignments in iD - which to use?

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Imagery variations/misalignments in iD - which to use?

John Willis
Another quick question:

Mapping in Japan.


in iD, we have some imagery to choose from:

-  Bing imagery (from very very good to meh),
- a Japanese government GIS render (which is okay, a patchwork of different vector illustrations) [Japan GSI standard Map]
- a Japanese “ortho” view, imagery with almost no off-angle tilt. [Japan GSI ortho imagery]

The issue I am facing is that, even after some adjustment of the angle of bing imagery, there seems to be some distortion. things don’t line up well between the Bing and ortho maps in some places, and are much closer in others.  a *lot* of the mapping aligns with the bing imagery, but there are areas of obvious 2-3m distortion in places (the road is wavy), but other areas of newer/clearer imagry align with the ortho imagry.

It is a difference of only a few meters, but with the insane complexity of even farming roads, a clash of a few meters between different mappers looks bad.

What do you guys do in such situations?
_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Imagery variations/misalignments in iD - which to use?

Graeme Fitzpatrick
I just go with the best available image for the area & follow that! Around here, that's almost always Bing as it's "usually" the clearest in our area (although you have to zoom in to 19 or 20 to get the best view :-()

Further out, it & the others are all so fuzzy that it's almost impossible to pick fine detail :-(

Thanks

Graeme


On Sat, 12 Jan 2019 at 10:26, John Willis <[hidden email]> wrote:
Another quick question:

Mapping in Japan.


in iD, we have some imagery to choose from:

-  Bing imagery (from very very good to meh),
- a Japanese government GIS render (which is okay, a patchwork of different vector illustrations) [Japan GSI standard Map]
- a Japanese “ortho” view, imagery with almost no off-angle tilt. [Japan GSI ortho imagery]

The issue I am facing is that, even after some adjustment of the angle of bing imagery, there seems to be some distortion. things don’t line up well between the Bing and ortho maps in some places, and are much closer in others.  a *lot* of the mapping aligns with the bing imagery, but there are areas of obvious 2-3m distortion in places (the road is wavy), but other areas of newer/clearer imagry align with the ortho imagry.

It is a difference of only a few meters, but with the insane complexity of even farming roads, a clash of a few meters between different mappers looks bad.

What do you guys do in such situations?
_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Imagery variations/misalignments in iD - which to use?

Eugene Alvin Villar
In reply to this post by John Willis
On Sat, Jan 12, 2019 at 8:26 AM John Willis <[hidden email]> wrote:
The issue I am facing is that, even after some adjustment of the angle of bing imagery, there seems to be some distortion. things don’t line up well between the Bing and ortho maps in some places, and are much closer in others.  a *lot* of the mapping aligns with the bing imagery, but there are areas of obvious 2-3m distortion in places (the road is wavy), but other areas of newer/clearer imagry align with the ortho imagry.

I also encounter this "wavy" roads in imagery. I think they are the result of improper orthorectification by the imagery provider. Satellite imagery is often off-nadir (not photographed straight down) so providers correct for differences in terrain elevation by rectifying them based on available elevation models. Unfortunately, many elevation models like SRTM cannot distinguish between buildings and hills and so roads are often distorted around tall buildings in many parts of the world.

I don't have any good solution for this aside from trying to get access to better imagery so I just try to map things as best as I can. It may also help to avoid micromapping unless you are sure that the imagery is really good.


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Imagery variations/misalignments in iD - which to use?

André Pirard-2
On 2019-01-12 06:33, Eugene Alvin Villar wrote:
On Sat, Jan 12, 2019 at 8:26 AM John Willis <[hidden email]> wrote:
The issue I am facing is that, even after some adjustment of the angle of bing imagery, there seems to be some distortion. things don’t line up well between the Bing and ortho maps in some places, and are much closer in others.  a *lot* of the mapping aligns with the bing imagery, but there are areas of obvious 2-3m distortion in places (the road is wavy), but other areas of newer/clearer imagry align with the ortho imagry.
Bing is sometimes (often? I almost never use it) inconsistent (offset) with itself at different zoom levels.
I also encounter this "wavy" roads in imagery. I think they are the result of improper orthorectification by the imagery provider. Satellite imagery is often off-nadir (not photographed straight down) so providers correct for differences in terrain elevation by rectifying them based on available elevation models. Unfortunately, many elevation models like SRTM cannot distinguish between buildings and hills and so roads are often distorted around tall buildings in many parts of the world.
I don't think that orthorectification uses elevation maps as correction. Rather, they combine shots from different angles to compute the position of objects and that computes their elevation. A nadir photograph of a building is uninteresting because it only shows the roof. Side shots show the walls and allow computation of their height and ground location.  Aerial imagery shows only one of the many shots they have.
I don't have any good solution for this aside from trying to get access to better imagery so I just try to map things as best as I can. It may also help to avoid micromapping unless you are sure that the imagery is really good.
The orthorectification we can use in Wallonia (PICC) is really perfect. Of the lines I see it computes in places I know, almost only the center of the road can be not wrong but debatable. Sometimes lanes that have been changed to parking or things like that are not up to date, but that's manual fixing, not orthorectification.

All the best,

André.


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Imagery variations/misalignments in iD - which to use?

voschix
Let me just inject that the wavy-roads problem is also present at Big G.
Go have a look at

Back to the argument of the thread: Generally speaking, of the maps we have available for OSM in Italy, the best ones are the oldest ones: PCN2006, poor resolution but near-perfect position (based on many GPX tracks) and good parallax correction. The worst are normally Bing. They have not only a position shift of many meters (that varies with the location), but also considerable problems with parallax correction, up to tens of meters (very poor parallax correction on hilly terrain). The others DigitalGlobe, Esri, Mapbox are in-between, but are have only shifted positions, but not the parallax problems of Bing.
Obviously I have no idea if Bing in Italy is technically the same as in Japan.

On Sat, 12 Jan 2019 at 18:41, André Pirard <[hidden email]> wrote:
On 2019-01-12 06:33, Eugene Alvin Villar wrote:
On Sat, Jan 12, 2019 at 8:26 AM John Willis <[hidden email]> wrote:
The issue I am facing is that, even after some adjustment of the angle of bing imagery, there seems to be some distortion. things don’t line up well between the Bing and ortho maps in some places, and are much closer in others.  a *lot* of the mapping aligns with the bing imagery, but there are areas of obvious 2-3m distortion in places (the road is wavy), but other areas of newer/clearer imagry align with the ortho imagry.
Bing is sometimes (often? I almost never use it) inconsistent (offset) with itself at different zoom levels.
I also encounter this "wavy" roads in imagery. I think they are the result of improper orthorectification by the imagery provider. Satellite imagery is often off-nadir (not photographed straight down) so providers correct for differences in terrain elevation by rectifying them based on available elevation models. Unfortunately, many elevation models like SRTM cannot distinguish between buildings and hills and so roads are often distorted around tall buildings in many parts of the world.
I don't think that orthorectification uses elevation maps as correction. Rather, they combine shots from different angles to compute the position of objects and that computes their elevation. A nadir photograph of a building is uninteresting because it only shows the roof. Side shots show the walls and allow computation of their height and ground location.  Aerial imagery shows only one of the many shots they have.
I don't have any good solution for this aside from trying to get access to better imagery so I just try to map things as best as I can. It may also help to avoid micromapping unless you are sure that the imagery is really good.
The orthorectification we can use in Wallonia (PICC) is really perfect. Of the lines I see it computes in places I know, almost only the center of the road can be not wrong but debatable. Sometimes lanes that have been changed to parking or things like that are not up to date, but that's manual fixing, not orthorectification.

All the best,

André.

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging