[OSM-dev] Coastline generalization tool and data

Christoph Hormann chris_hormann at gmx.de
Tue Feb 19 12:04:19 GMT 2013


On Tuesday 19 February 2013, Michal Migurski wrote:
> Christoph, this is so cool. Excellent work!

Thanks.

> I've been hammering away at an OSM generalization effort of my own,
> focused on the global preparation of simplified linework for major
> roads and route relations:
> http://www.openstreetmap.us/~migurski/streets-and-routes/

For generalizing roads I think there are four major components:

1. removing smaller roads where there is no room to show them properly
2. summarize dense networks of roads of similar size
3. move roads to be able to distinctly show them without overlap with 
other features or themselves
4. simplify the individual road segments

From your description I have the impression that a major part of your 
effort is in 2. although I could not clearly identify cases of 
summarization in your examples.

From the samples it seems your technique sometimes produces gaps between 
roads which are in fact connected.  Apart from that it looks quite 
good - for curved roads the simplification seems a bit strong unless 
you mean to use some kind of smooth spline rendering.

By the way for the coastlines I also looked into the idea of using 
vector skeletons but I soon realized this would be prohibitively slow 
on a global scale.  From the pure program design point this would still 
be an interesting idea.

> I did find one weird part of your data, around Boston where the
> peninsula appears detached from the mainland.

That's another canal tagged as coastline:

http://www.openstreetmap.org/browse/way/22721484

-- 
Christoph Hormann
http://www.imagico.de/



More information about the dev mailing list