[OSM-dev] OT: OGR, Douglas-Peucker

[Lars Aronsson]

Erik Johansson wrote:

> Implementing Douglas-Peucker isn't that hard and there are some 
> pieces of example code around, but I failed hard in using it. 
> Not because it doesn't work, it does and it's really great for 
> lines. But I can't reallt figure out how to use it with graphs, 
> when two lines are connected I don't know what to remove.

I think I don't really understand the issue you describe.

My perspective as a newcomer to this field is this: If I want to 
dig deeper into how I can use the free vector map data we get out 
of OSM, should I start with OGR and implement my own 
Douglas-Peucker or should I go for some other (which?) open source 
toolbox where such algorithms are already implemented?

Anyhow, I assume one application could be this: If you are drawing 
the OSM map in a smaller scale (something like zoom=7), and if the 
rendering takes too much time, you could let Douglas-Peucker 
reduce the amount of data by removing several nodes that connect 
only two line segments, but keep all nodes that connect only one 
line segment (end points) or three or more (junctions). This would 
simplify the curves/lines for long country roads.  For cities with 
dense patterns of 3- and 4-connecting nodes, some other kind of 
data reduction could be used.  I'm just guessing.

Cartographers use the word "generalization" for this kind of data 
reduction.  Doesn't Mapserver have any automatic support for this?

The OSM dataset might be optimal for rendering in scale 1:30,000 
(zoom=13) and the necessary generalization for scale 1:300,000 
(zoom=10) or 1:3,000,000 (zoom=7) could perhaps be run as a 
nightly batch conversion.  The on-the-fly tile generation could 
then use these reduced datasets for the reduced scales, keeping 
the complexity (number of nodes) per tile at a constant level.

Again, these are just my loose fantasies, based on no experience 
at all.  My immediate question is a search for a useful toolbox.


-- 
  Lars Aronsson (lars at aronsson.se)
  Aronsson Datateknik - http://aronsson.se