On Tuesday 06 April 2010, TimSC wrote: > Regarding the technical difficulty, it remains to be seen if automatic > tracing is possible and of sufficient quality. But it is presumptuous to > assume it is impossible at this stage.
I want to put my oar in for finding a consensus that we will get the best consistent results from automatic tracing of the raster. I don't think it is a particularly hard problem. From the looks of it (take these numbers with a pinch of salt), I'd say about 95% of buildings are rectangles. About 99% of buildings are purely orthogonal lines. We can gain a lot from hunting for corners and snapping to orthogonals. This is a vastly simplified subset of the raster to vector problem. My preferred solution would be doing a country-wide generation of a buildings.osm which users can merge from. Hence no need for special plugins or tools to be distributed to regular contributors. The tile boundary situation is still an open problem though. Unless someone has a machine with terabytes of memory ;)? robert. _______________________________________________ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
