[OSM-dev] Simplifying ways

[BG]

is there a possibilty to simplify without douglas peucker?
i tried simplifying with postgis and i didn't like the result - some 
ways were doubled



On 5/8/2012 9:14 AM, Paul Norman wrote:
>> From: Paul Norman [mailto:penorman at mac.com]
>> Subject: [OSM-dev] Simplifying ways
>>
>> I'm working with a process where I generate a .osm file and need to
>> simplify the ways in output on the command line. JOSM's simplify way
>> command isn't an option as it requires interaction and it is also fairly
>> slow on large files.
>> Using ogr2ogr to simplify the source files is not an option as it could
>> cause the different layers to no longer agree with each other.
>>
>> I'd rather not re-implement Douglas-Peucker if someone has already done
>> it.
> To summarize the options given to me through email and IRC there are
>
> 1. JTS[1]. This is not OSM-specific and written in Java. I would need to
> write code to convert OSM data to the Geometries.
>
> 2. gdal Simplify[2]. This would be the easiest to integrate into ogr2osm
> except that I need to simplify after conversion to OSM objects and duplicate
> node removal or two ways that share nodes for part of their length. There
> are also cases where an optimal solution would not be arrived at since
> points that were removed could be re-added by another object.
>
> 3. osmosis-simplifyways[3]. This is based on JTS and uses osmosis.
> Unfortunately osmosis rejects .osm files with negative IDs so I would need
> to either multiply all IDs by -1 before and after or patch osmosis.
>
> 4. Schuyler Erle's PD implementation[4] of DP in python. This is a
> non-recursive implementation of DP. I would have to write code to read the
> .osm file, identify points that would be kept anyways, reproject to a
> projection where distance can be measured and them simplify.
>
> 5. scanaerial's GPL implementation[5] of DP in python. This is a recursive
> implementation. The same issues apply as with option 4.
>
> Given the choices, I think I'm going to end up implementing something based
> on 4 or 5. I want something that will preserve any points in a relation or
> where another way meets the way being simplified but does simplify when two
> ways share nodes for a stretch.
>
>
> [1]: http://tsusiatsoftware.net/jts/main.html
>
> [2]: http://gdal.org/python/osgeo.ogr.Geometry-class.html#Simplify
>
> [3]: https://github.com/podolsir/osmosis-simplifyways
>
> [4]: http://mappinghacks.com/code/dp.py.txt
>
> [5]:
> https://github.com/jonasstein/scanaerial/blob/master/scanaerial_functions.py
> #L29
>
>
>
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/dev
>