[postgis-users] How to eliminate small gaps produced by ST_union?

[Alexandre Neto]

I'm trying to Aggregate\Dissolve a set of adjacent, but not touching polygons in Postgis. The geometries were created editing the Postgis table in QGIS created, with snapping and avoid overlap in the layer.

I used ST_Union like this:

SELECT ST_Union(the_geom)

FROM table;

Small gaps similar to lines appear in the result. This happens when the adjacent polygons do not share the exact same vertices.

Is there a way to eliminate this "gaps" or "small angles" within Postgis?

I know that v.clean in GRASS can eliminate small angles, and correct topology, but that would mean I needed to copy the features to GRASS, correct them and copy it back to Postgis.

Thanks in advance,

Alexandre Neto 

[Sandro Santilli]

You may try to construct a topology within PostGIS-2.0, using
a small tolerance, and then get the geometries back.

--strk;

_______________________________________________
postgis-users mailing list
postgi...@postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-users

[pcr...@pcreso.com]

My impression is that working topologies is slow, especially with large datasets - lots of computation involved. How effective would this be with hundreds of thousands or millions or small polygons? Just running ST_Union() takes hours to days. Thanks Brent Wood --- On Fri, 10/26/12, Sandro Santilli <st...@keybit.net> wrote:

[Bo Victor Thomsen]

Try to use function st_cleanpolygon or st_buffer with a radius of 0.0.

Regards
Bo Victor Thomsen
Aestas-GIS
Denmark

[Paul Ramsey]

If you don't mind altering your other edges somewhat in the process,
dilate/erode will blow those away,

http://blog.opengeo.org/2010/11/22/removing-complexities/

P.

On Thu, Oct 25, 2012 at 10:44 AM, Bo Victor Thomsen
<bo.victo...@gmail.com> wrote:
>
> Try to use function st_cleanpolygon or st_buffer with a radius of 0.0.
>
> Regards
> Bo Victor Thomsen
> Aestas-GIS
> Denmark
>
> Den 25-10-2012 18:54, Alexandre Neto skrev:
>
> I'm trying to Aggregate\Dissolve a set of adjacent, but not touching polygons in Postgis. The geometries were created editing the Postgis table in QGIS created, with snapping and avoid overlap in the layer.
>
> I used ST_Union like this:
>
> SELECT ST_Union(the_geom)
> FROM table;
>
> Small gaps similar to lines appear in the result. This happens when the adjacent polygons do not share the exact same vertices.
>
>

[Sandro Santilli]

It will be _very_ slow.
So much that you may prefer to make the topology persistent.

Another approach I'd like to see implemented would be to expose a tolerance
to the GEOS implemented functions. Basically exposing the GEOS
PrecisionModel. By doing that you could specify a tolerance in the overlay
operations and the result should never have gaps bigger than the given
tolerance.

--strk;

--

http://www.cartodb.com - Map, analyze and build applications with your data

~~ http://strk.keybit.net

[Dave Potts]

Have you considered curve smoothing or apply a small buffer around the
orginal data/

--

[Alexandre Neto]

Thank you for all the answers.

I did not try the topology, as I need to study its implementation a little better.

I did not found the st_cleanpolygon function and using (St_Buffer, 0) did not worked:

The dilate\erode visually corrects the problem, but of course introduce a small rounding in the angle introducing a small overlaping with other adjacent polygons...

I guess that the tolerance with the GEOS PrecisionModel would be nice, I think that is the method used by ArcGIS products to "insure" geometry "topology" after aggregation operations.

Any more ideas?

Alexandre Neto

[Sandro Santilli]

On Fri, Oct 26, 2012 at 10:05:29AM +0100, Alexandre Neto wrote:
> Thank you for all the answers.
>
> I did not try the topology, as I need to study its implementation a little
> better.
>
> I did not found the st_cleanpolygon function and using (St_Buffer, 0) did
> not worked:
>
> The dilate\erode visually corrects the problem, but of course introduce a
> small rounding in the angle introducing a small overlaping with other
> adjacent polygons...
>
> I guess that the tolerance with the GEOS PrecisionModel would be nice, I
> think that is the method used by ArcGIS products to "insure" geometry
> "topology" after aggregation operations.
>
> Any more ideas?

You may try collecting all vertices of input into a MULTIPOINT, then
ST_SnapToGrid that MULTIPOINT, then ST_Snap each of the input geometries
to the resulting MULTIPOINT, with your desired tolerance.
Finally pass the result trough ST_MakeValid to drop collapsed polygons.

Should work fine. Let us know.

--strk;

[Alexandre Neto]

That seamed to have worked.

Here is my implementation of Sandro guidelines:

SELECT ST_Union(v.the_geom)

FROM

(SELECT ST_MakeValid(St_Snap(t.the_geom, 

        (SELECT ST_SnapToGrid(st_collect(f.the_geom), 0.001) as the_geom

         FROM (SELECT (ST_DumpPoints(g.the_geom)).geom as the_geom

               FROM table1 as g) as f),

        0.001)) as the_geom

FROM

    table1 as t) as v;

It's obviously a slow process as it implies 2 sequential reads in the table, it takes 11ms with a very simple table containing only 2 polygons, as it implies 2 sequential reads in the table.

With a bigger table ( 701 polygons, 113147 vertices), it takes 762575 ms...

Since the problem occurs when I do the input of new polygons, I will try to use this procedures only for the new polygons and the adjacent ones.

Thank you for your help.

Alexandre Neto