[OSM-dev] A SOMMER CURIOSITY

Sandor Seres sandors39 at gmail.com
Wed Jul 22 20:02:18 UTC 2015


Let us move the focus from the high to the low/start end of the map-making
development for a short while. Yes, I know, the work and the issues at this
end are much less attractive but sometimes we just have to be there too. So,
here are some facts ("facts" underlined) from the last OSM dump, at this
time related to the missing holes in water areas (you may call them
islands). These "facts" may have changed the last day or two but this is
ignorable. Did you know that:

-There are 883 fake/wrong land areas in the coastline based planet_land?
These are land areas inside larger land areas (overlapped), mainly remaining
holes when people moved out coastline water objects to lakes and rivers. Of
these 427 are missing holes in the river and lake area objects and you are
never seeing them in maps. 

-There are still over 7000 active natural=land objects. Of these, 1275 are
still missing holes/islands in the planet_land water objects, lakes and
rivers.

-There are large number of overlaps between the water area objects in the
planet_land and lakes and rivers. These overlaps are in all possible
combinations. When two water objects (partly) overlap a hole in the overlap
in one but not in the other one is never visible (eventually, you will see
two shades of blue). There are 3953 such invisible holes/islands.

So, totally we have 6655 missing islands in most of the OSM maps at present.
The consequences are pretty well known for the map makers. There are several
ways/options how developers may compensate or correct these errors. But that
is another story. Of course, the best should be to correct these errors in
the OSM source data.

Finally, before I go, just a hint to researchers, higher grade students and
academics. The mentioned "missing holes/islands" issue hides several (many)
serious and complex themes in Polygon/Image Algebra, Algorithms and
Programing. Without pretending to give strong and formal problem
definitions, here are some of the possible themes:

-In the coastline data detect the wrong/fake areas, find where they belong
and distribute them.

-Some of the water area objects in planet_land and planest_lakes fully, or
partly, overlap. Find all hole/island objects in overlaps that are
replicated in both classes (or the contrary).

-In a given set of simple areas (one outer and arbitrary number of inner
borders), areas may have arbitrary positions/relations (no assumption about
mutual relations). Find the minimum simple area coverage (a set of
disjunctive simple areas that perfectly covers the areas in the former set).

Just to mention some.

Regards, Sandor.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20150722/b019aa63/attachment.html>


More information about the dev mailing list