[HOT] Helping Japan by tracing

Hans van Wijk hvwijk at planet.nl
Tue Mar 15 17:51:04 GMT 2011


> > the coastline in Japan is coming from the Japanese governement and 
> > tends to be pretty good. It is the reason why they are both 
> > natural=coastline and boundary=administrative.
> 
> If the tsunami has significantly _altered_ coastline, that'd 
> be a reason to change the natural tag.  I haven't seen any 
> yet in the post-quake imagery.

I was talking about pre-quake bing imagery, but somebody now started to
unglue the natural=coastline from the boundary=administrative in the places
that needed important change (i.e. harbour walls) and redraw the
natural=coastline. I think that's a good solution.

> > MLIT is one of the Japanese ministries which has tons of open Data.
> > Some of the data can be pretty old some can be pretty recent. I 
> > suspect you can alter them if you feel that the imagery 
> represents the 
> > ground truth, but be aware that some of those forests are huge.
> 
> The forests look like they were drawn from low-res sources 
> and interpolated to smooth them out.  Certainly there are 
> lots of cases where they don't reflect the ground truth even 
> approximately, which may be down to the passage of time, and 
> may also be a photogeographer or algorithm getting confused 
> between forest and farmland.

I'm guessing a combination of confused algorithms and poor alignment. The
outlines are coarse and the different patches are all misaligned by the same
amount.

Somebody in this thread also questioned the alignment of the pre-quake bing
imagery, but I believe that to be correct, judged by some GPS tracks that
are available for the region.

Thanks everybody for the answers!




More information about the HOT mailing list