[Talk-us] MapRoulette new challenge: connectivity

James Umbanhowar jumbanho at gmail.com
Tue Oct 30 14:37:47 GMT 2012


On Mon, 2012-10-29 at 15:25 -0600, Martijn van Exel wrote: 
> On Mon, Oct 29, 2012 at 2:43 PM, Ivan Komarov <jkomarov at gmail.com> wrote:
> > On Mon, Oct 29, 2012 at 3:33 PM, Frederik Ramm <frederik at remote.org> wrote:
> >> If one road ends near another road, that might actually be for a reason, and
> >> what looks like a shadow on the aerial image is in fact a fence - or the
> >> aerial image is outdated...
> >
> > That's true. But in the US disconnected roads are produced by buggy
> > TIGER import in 99% cases, so they definitely need fixing. On the
> > other hand I have to admit that attentiveness required, as once I
> > screwed Chicago downtown not paying enough attention to road levels.
> > But in countryside these remote fixes are rather safe, I guess.
> >
> 
> That is right. I forgot to mention that maybe 70% of the connectivity
> errors (wild guess) is the result of TIGER being county-based, leading
> to 1) lots of connectivity errors at county boundaries, and 2) wildly
> varying quality between counties, even within states some say (I
> haven't really looked into that in any detail.)

I just wanted to add a couple points here in regards the nature of
connectivity issues.  First, I think that Martin's wild guess is off.
I've done about 20 of these (which is getting close to a sufficient
sample) and I would say that about 70% of the errors are due to human
mappers.  I've had 2 that have been impossible to tell based on Bing
imagery, but were solvable using Tiger 2012 data.  

So in regards to whether this is a US specific game, I would say that
the problems are not, but the presence of high quality aerial imagery
and decent quality road layers are.  Also, Martin, would it be possible
to direct folks to instructions on how to load Tiger data into their
editors, i.e. https://wiki.openstreetmap.org/wiki/TIGER_2012?

James




More information about the Talk-us mailing list