On Tue, Jul 6, 2010 at 2:42 PM, Jim Morgan <j...@datalude.com> wrote:
> Eugene Alvin Villar wrote, On Tuesday, 06 July, 2010 11:57 AM: > > To clarify things, the sawtooth detection script is quite naive. It > > simply detects if there are at three or more series of nodes where each > > pair of adjacent nodes have the same latitude or longitude. This will > > also detect any three linear nodes that all have the same latitude or > > longitude like this: o----o----o > > There are a number of cases where a near-straight line is acceptable. Maybe > it would be better -- and I'm not sure if this is possible -- to examine, > say, a series of three nodes. It would check if the first two have the same > lat or long. If they have the same lat, then the second and third points > would need the same long; if they have the same long, then the second and > third points would need the same lat. Then you'd be correctly identifying > the step-fashion jaggies, rather than straight lines. > > > To increase certainty, you could make this a series of four, or five > points. Again I don't know if this is possible or plausible, but it would > seem like a better pattern to look for. Not sure if the formatting will come > through but .... > > > p1 |_____ p2 > | > | > p3 |______ p4 > | > I intentionally wanted to detect collinear nodes since I wasn't sure if the original SRTM-based data have those collinear nodes or not. In any case, the script detects sawtooth coasts if the latitude or the longitude is *exactly* the same, right down to the 7th decimal place (which translates to an accuracy of about 1 cm). So if there are a series of coastline nodes that have the same latitude or longitude for each adjacent pair of nodes, then they are most likely generated from raster data, like SRTM. I don't think Mother Nature created coasts that follow latitudes and longitudes. :-)
_______________________________________________ talk-ph mailing list talk-ph@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ph