You are not logged in.

#1 2015-07-06 17:28:05

Fernando Trebien
Member
From: Porto Alegre, Brazil
Registered: 2013-05-18
Posts: 888

Angular precision of curves: how much is enough?

If you were about to seriously represent a small (say 6 meters wide) circular man-made structure in OSM, how would you do it? Several options:

1. Point, because it is not large enough to be an area.
2. Square, because at that scale shape detail is not so important, most people just need to know its location and approximate size. Maybe also because that's what JOSM's Simplify Way action does for features of very small size.
3. Hexagon, because it is enough to tell the users it is not rectangular like most buildings. Implies 60° is the recommended minimum angular precision for most curves.
4. Octagon, to avoid confusion with hexagonal features. Implies 45° is the recommended minimum angular precision for most curves.
5. Dodecagon, because octagon is not enough. Implies 30° is the recommended minimum angular precision for most curves.
6. Heptadecagon, because it is a known cultural threshold between regular polygons and circles. Implies 21° is the recommended minimum angular precision for most curves.
7. Octadecagon, because it is above the heptadecagon threshold and even, so its edges align with rectangular features which may lie nearby. Implies 20° is the recommended minimum angular precision for most curves.
8. Hexadecagon, because it is close to the heptadecagon and easier to subdivide than the octadecagon and also even. Implies 22.5° is the recommended minimum angular precision for most curves.
9. Enneadecagon, because that's what iD's Create circle function does regardless of circle size. Implies 19° is the recommended minimum angular precision for most curves.
10. More than 19 edges.
11. Less than 19 edges, but other number.

And what would you answer if the curve represented a natural feature such as a coastline or the contour of a forest?

Some famous such objects already mapped and similar ones in the same country:

- Pisa Tower: 16 m wide, 23 edges (16°)
- hut in a park, Rome: 7 m wide, 8 edges (45°)

- BT Tower, London: 13 m wide, 19 edges (19°)
- nearby fountain: 7 m wide, 8 edges (45°)

- Round Tower, Copenhagen: 17 m wide, 19 edges (19°)
- nearby hut: 11 m wide, 9 edges (40°)

- Fallturm, Bremen: 9 m wide, 18 edges (20°)
- a small pavillion, Berlin: 8 m wide, 8 edges (45°)

More examples welcomed.

While crude geometry is undesirable, I also understand that there is such a thing as excessive detail (that which does not add interesting information), so there has to be something like an approximate ideal level of detail that balances the need for precise information and maintenance effort. Here are some links to support this view:

Last edited by ftrebien (2018-03-13 21:03:57)

Offline

#2 2015-07-07 11:11:16

Tordanik
Moderator
From: Germany
Registered: 2008-06-17
Posts: 2,807
Website

Re: Angular precision of curves: how much is enough?

Whether I would consider using a node depends on the type of feature. I would never use a node for a building, for example.

Personally, I would use at least draw 12 way segments, but likely more. But I really wish there was some way to make it unambiguous, though, like tagging it round=yes or something.


OSM in 3D: OSM2World

Offline

#3 2018-01-12 18:27:40

ftrebien
Member
From: Porto Alegre, Brazil
Registered: 2013-05-18
Posts: 888

Re: Angular precision of curves: how much is enough?

I didn't notice but there is one more view expressed in the wiki here. So far the average of three opinions suggests an octadecagon (20°), quite close to iD's Create circle function.

I would prefer a standard number of edges divisible by 4 since the majority of roundabouts connect roads aligned at straight angles. That would be an icosagon (18°).

Last edited by ftrebien (2018-01-18 16:55:08)

Offline

Board footer

Powered by FluxBB