[OSM-dev] Doing street names from aerial imagery

[John Smith]

On 24 March 2010 21:31, John Robert Peterson <jrp.crs at gmail.com> wrote:
> Are you willing to accept that if a road changes name part way though, and
> that missing this is a reasonable error?

What is the probability of this occurring? I'm guessing fairly low,
although you could minimise this by targeting the intersections
closest to the start and end nodes of a way, I don't think every
single intersection along the way is worth the effort.

> Is it reasonable to assume that all volunteers are especially the same (they
> all have a car, and have no specific preferences)?

This could be handled some what by using current cake method to break
things up, but with a slight twist where the boundaries can be
flexible to make routing more efficient.

> Is it reasonable to assume that everyone starts and finishes at a
> predetermined "base"?

I'd assume a fuzzy cake area, rather than a shared start point.

> Is it reasonable ot assume that if someone finds a 1 way system, or other
> restriction, that the route will go wrong, and this is ok?

I think the intersections are more important than the way itself, any
ways that are one way you might only need to use the start
intersection.

> How large an area are we dealing with here? most algorithms for this type of

I'd assume a suburb worth, any smaller and it wouldn't be worth using
complex routing, any larger and it may not be worth the effort due to
inefficients as you point out, depends on the number of volunteers I
guess.

> In short: I'd recommend just using intuition, but if that's not an option,
> give a load more details, and we can work something out.

I was after a general solution, I think it would be useful for OSM...