Is it possible to convert Polygon to centerline linestring?

Question

I have a Polygon which I want to convert to centerline linestring.

IMAGE 1

I used the Vector to convert to Polyline but they are not center line as under:

VECTOR TOOL

IMAGE 2

Polyline created by the above tool is not centerline polyline.

IMAGE 3

What I am expecting to get is as under (centerline per road, instead of the lines on the side of the road[Image 3]):

IMAGE 4

Answer 1

Answer: You can.

I found time for you and hurry to help you. My answer is now practical.

The original data and table names are fictitious, do not forget to rename them with reference to your example.

Sit in my car and go together to the stop - the "center line of the road."

The initial geodata are polygons of a road network with a width of about 11 m in the EPSG: 4326 coordinate system, broken into segments, see Fig. 1.

1. In a convenient way for you, create a negative buffer, for example in pgAdmin:

create table buf_negative as SELECT ST_Buffer (geom, -0.000045) as geom FROM buf_dump; or in QGIS tools: "Vector-Geoprocessing-Buffer Zones". Thus, we reduced the width of the roads from 11 m to 1 m, see Fig.2

2. Convert the negative buffer (1 m) into lines, for example with QGIS tools: "Vector-Processing geometry-polygons in a line" result see Fig. 3

3. Extract from the lines of the point, for example with the QGIS tools: "Vector-Processing geometry-Extract nodes", the result see Fig. 4

4. Create point buffers from them, so that they intersect each other, Further combine them with each other using the tool "Vector>Geoprocessing>Merge" and extract centroids from adjacent buffers, the result is shown in Fig. 5 (centroids are red dots)

5. From the centroid points, create the Delaunay Triangulation, for example with the QGIS tools: "Vector-Processing geometry-Delaunay triangulation", and trim the result with your 1 m negative buffer polygons, see the result of Fig. 6

6. Blast the lines into segments, for example, using QGIS Desktop with GRASS tools: "Data Analysis - Lines-Explode Lines", see Fig. 7

7. Run the following script in PGAdmin:

create table buf_centre_ln as SELECT ST_Intersection (a.geom, b.geom) AS geom FROM explode_ln as a JOIN buf_negative as b ON ST_Within (a.geom, b.geom);

see the result in Fig. 8

9. Remove duplicating lines and do with the center lines, whatever,

Answer 2

The library pygeoops also has an implementation to calculate a centerline of a polygon: pygeoops.centerline. It also removes short branches, simplifies the result,... to get a "clean" result.

Sample script:

import pygeoops
import shapely
polygon = shapely.from_wkt("POLYGON ((0 0, 0 8, -2 10, 4 10, 2 8, 2 2, 10 2, 10 0, 0 0))")
centerline = pygeoops.centerline(polygon)

Result:

Disclaimer: I'm the developer of pygeoops

Answer 3

It is possible to achieve it, on PostgresGis, with one of the options of the free tool "geom8tricks" With the "PolygonsToLines" option, once you select the origin polygons and the destination Database connection has been correctly configured, you get the line objects, centered on the polygons . The program does it in two stages, in a first one it obtains the decomposition of the polygon in lines, by means of st_skeleton function

In a second stage, the longest line of all those that make up the object is selected

The problem of intersections is solved whenever there is a primary key column in the origin polygons, which allows to distinguish between the objects that intersect

Answer 4

I looked for the Geom8tricks suggestion above and found this link which seems to also include the PolygonToLines feature. I haven't used it myself.

For anyone still looking for this I think he's referring to http://geom8tricks.com/index_ml.html?lang=en

Answer 5

There is a python library centerline 1.0.1.

https://centerline.readthedocs.io/en/latest/

https://github.com/fitodic/centerline/issues

On github for the lib centerline (issues) user Turtle6665 proposed an alternative solution for shapely 2.0 and geopandas.

import shapely
import geopandas as gpd
def construct_centerline(input_geometry, interpolation_distance=0.5):
   #find the voronoi verticies (equivalent to Centerline._get_voronoi_vertices_and_ridges())
   borders = input_geometry.segmentize(interpolation_distance) #To have smaler verticies (equivalent to Centerline._get_densified_borders())
   voronoied = shapely.voronoi_polygons(borders,only_edges=True) #equivalent to the scipy.spatial.Voronoi
#to select only the linestring within the input geometry (equivalent to Centerline._linestring_is_within_input_geometry)
   centerlines = gpd.GeoDataFrame(geometry=gpd.GeoSeries(voronoied.geoms)).sjoin(gpd.GeoDataFrame(geometry=gpd.GeoSeries(input_geometry)),predicate="within")
   return centerlines.unary_union
centerlines = construct_centerline(open_area_Road[0],interpolation_distance=0.5)