Combine Multiple Polygons into a Single Polygon with Holes Using Python

geopandaspolygonpythonshapely

I have the following GeoDataFrame with multiple polygons which represents a single area (raster_val = 1.0) with holes in the middle (raster_val = NaN):

                                       geometry           raster_val
0   POLYGON ((8.77500 47.70000, 8.77500 47.67500, ...         NaN
1   POLYGON ((8.90000 47.40000, 8.90000 47.30000, ...         NaN
...
9   POLYGON ((6.12500 46.27500, 6.12500 46.25000, ...         1.0
10  POLYGON ((8.57500 47.80000, 8.57500 47.77500, ...         1.0
11  POLYGON ((8.90000 45.85000, 8.90000 45.82500, ...         1.0

I would like to combine polygons corresponding to raster_val = 1.0 into a single Polygon/Multipolygon but preserving the holes indicated by the polygons with raster_val = NaN.

I've already tried geopandas dissolve() method:

polygons = polygons.dissolve(by='raster_val')

This combines raster_val = 1.0 polygons but it ignores all raster_val = NaN polygons so the combined multipolygon doesn't have any holes. Do you have any suggestions on how to include those NaN holes?

Best Answer

Perhaps shapely's unary_union will help, where df is your GeoDataFrame:

import pandas as pd
from shapely.ops import unary_union

hole_indx = pd.isna(df.raster_val)

holes = unary_union(df.loc[hole_indx, 'geometry'])
polys = unary_union(df.loc[~hole_indx, 'geometry'])

polys_and_holes = polys.difference(holes)

Related Solutions

Convert Multiple Polygons and MultiPolygons into Single MultiPolygon in GeoJSON

You could do this using Javascript Topology Suite which will work with Node.js. Start with an empty MultiPolygon (or the first geometry in your collection) and union this with each (Multi)Polygon in your collection. You can only have one format for the whole collection, obviously, as properties are one to one with the geometry in the GeoJSON. Here are some jsts/node.js examples.

There is also a GeoJSON parser in jsts, which will return a jsts geometry. So putting it all together, you would do something along the lines of:

var jsts = require("jsts");

var reader = new jsts.io.GeoJSONParser();

//read your geometries
var geoms = reader.read(geojson);

//grab the first one
var multipolygon = geoms[0];

//union with all the others
for (var x=1; x < geoms.length ; x++){
    multipolygon = multipolygon.union(geoms[x]);
}

This is untested, but the general idea should work. You would have to recreate the GeoJSON with properties afterwards, but this should be easy with the built in JSON methods in Javascript.

[GIS] Convert contours to polygons maintaining attributes using Python

A simpler way would be to use geopandas and shapely:

 import geopandas as gpd
 import shapely.geometry as geom
 def polyline_to_polygone(input_file, output_file):
     gdf = gpd.read_file(input_file)
     gdf['geometry'] = gdf['geometry'].apply(lambda x: geom.Polygon(x.coords))
     # in order to have donut geometries, subset polygons from each others
     for index in list(gdf.index):
           indices = list(gdf.index)
           indices.remove(index)
           for index_ in indices:
                try:
                    gdf.loc[index, 'geometry'] = gdf.loc[index, 'geometry'].difference(gdf.loc[index_, 'geometry'])
                except ValueError:
                    pass
     gdf.to_file(output_file)

Best Answer

Related Solutions

Convert Multiple Polygons and MultiPolygons into Single MultiPolygon in GeoJSON

[GIS] Convert contours to polygons maintaining attributes using Python

Related Question