[OSM-talk] (magical?) road detector

Steve Bennett stevagewp at gmail.com
Sun Feb 6 10:18:50 GMT 2011

On Sun, Feb 6, 2011 at 1:01 PM, Thibault North <tnorth at fedoraproject.org> wrote:
> In the mapping process (with JOSM or such tool), following roads is not really
> a problem, especially when they are not too sinuous (and that's when the road
> detector works well...). It can be done in a few clicks. Maybe the tool should
> try and act differently (but that is more GUI/UI related), and we could imagine
> the following scenario:
> - The user wants to map roads and selects a "road extraction" tool.
> - He roughly follows a road, maintaining a click (as you would do to paint
> with a brush in image processing softwares)
> - The algorithm knows the approximate path, and tries to fit exactly the center
> of the road.

I think this would be the best way to do it. If an editor could
perform each of the following operations with a single click or
- draw straight way segment from the last node to the mouse cursor
- draw from the last node to the mouse cursor, following a perceived road
- undo last automatically drawn section

Or perhaps even the following:
- advance from the last node a further X distance, following roads
(where X is dependent on zoom level)

...then you have the makings of a very efficient process for tracing
roads off imagery. The last operation above would let you keep hitting
a key to advance a road until something goes wrong, then backtrack and
fix it manually.

What I saw in my testing was that most of the time (perhaps 70%) it
got the road right, and sometimes it was just hopelessly wrong. That
is presumably because the algorithm is determined that there *is* a
road to find. It would be better for it to give up and not draw a road
at all if its confidence isn't high.


More information about the talk mailing list