[talk-ph] Automated assessment of OSM coastline data quality

Eugene Alvin Villar seav80 at gmail.com
Mon Apr 15 14:29:19 UTC 2013


Well Landsat's resolution is low so any coastline traced using Landsat
(like almost all of the coastline corrections 3 years ago) will have a low
average distance between nodes marking it as low to average quality. So
while the offset may be a problem, I think the low node density can be
already used as a flag for coastlines that still need improvement.


On Mon, Apr 15, 2013 at 10:28 AM, maning sambale <emmanuel.sambale at gmail.com
> wrote:

> Interesting.  Thanks for sharing Eugene.
>
> I think the common issue for the Philippines is the third case in the
> first image ("coastline with detailed but inaccurate approximation").
> When we started the coastline correction 3 years ago, there was very
> little hi-res imagery.  For my part, I did most of the corrections
> using the Landsat imagery.  Vewing with bing's hires imagery, there is
> an obvious offset from the Landsat derived correction.  As mentioned
> in the article, this is not considered in the output visualization.
>
> If we can get Cristoph's code for making this visualization, maybe we
> can setup another webmap (like the Sawtooth Coastlines webmap) to
> coordinate improvements?
> Helping Indonesia improve its coastline is a good inter-community
> project as well. :)
>
>
> On Sun, Apr 14, 2013 at 3:26 AM, Eugene Alvin Villar <seav80 at gmail.com>
> wrote:
> > Hi guys,
> >
> > Here's a nice analysis done by Christoph Hormann in automated assessment
> of
> > the quality of OSM's coastline data:
> > http://www.imagico.de/map/coastline_quality_en.php
> >
> > If you don't want to read the whole article, here's the TL;DR version:
> >
> >> Coastlines where there are a few nodes per kilometer AND where the angle
> >> at each node is large means that the coastline is most likely to be of
> low
> >> quality. The last map image in the article highlights areas in the world
> >> where these "low" quality coastlines are located such as large parts of
> >> Indonesia.
> >
> >
> > The Philippines is on the "average" to "probably good" side of the
> scale. I
> > think this is thanks in a large part to the coastlines correction
> > collaboration the OSMPH community did 3 years ago:
> > http://wiki.openstreetmap.org/wiki/Philippines/Coastline_corrections
> >
> > But if you look at Japan, you can see that it is on the "probably very
> good"
> > side of the quality scale. So the Philippines still needs a lot of
> > improvement.
> >
> > If you want to help out improving the OSMPH coastlines, you can do the
> > following:
> > 1. Go to the OSMPH imagery coverage map (either
> > http://maning.github.io/Imagery_Coverage_Map/ or
> > http://tools.openstreetmap.org.ph/imagery_coverage/ )
> > 2. Look for remote/provincial coastline areas with high-res Bing
> satellite
> > imagery
> > 3. Zoom into a coastline, then switch between the Bing and OSM/MapQuest
> > background layers to check the coastline accuracy
> > 4. Click on the "Edit in Potlatch" or "Edit in JOSM" link to edit away!
> (You
> > need to enable JOSM's Remote Control feature [it is enabled by default in
> > newer JOSM versions] for the "Edit in JOSM" link to work.)
> >
> > If you're using JOSM, the Improve Way Accuracy tool
> > (http://josm.openstreetmap.de/wiki/Help/Action/ImproveWayAccuracy) is
> > *extremely* helpful. If will eliminate the need to click-drag your mouse.
> >
> > Maybe in the future, we can recreate the quality algorithm developed by
> > Christoph Hormann and create a webmap specific for the Philippines. :-)
> >
> > Keep on mapping!
> >
> > Eugene
> >
> >
> > _______________________________________________
> > talk-ph mailing list
> > talk-ph at openstreetmap.org
> > http://lists.openstreetmap.org/listinfo/talk-ph
> >
>
>
>
> --
> cheers,
> maning
> ------------------------------------------------------
> "Freedom is still the most radical idea of all" -N.Branden
> wiki: http://esambale.wikispaces.com/
> blog: http://epsg4253.wordpress.com/
> ------------------------------------------------------
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-ph/attachments/20130415/d0fa9f07/attachment.html>


More information about the talk-ph mailing list