有没有办法在 python 中使用 shapely 和 geopandas 找到 3 条或更多线相互接触的点坐标列表？

Question

I have a Geopandas DataFrame with each row containing a shapely LineString. I need to find a list of points where three or more LineStrings touch each other. For example, in figure Input image , I need the coordinates of the point where the three colored lines meet.

After I find all the points where three lines intersect, I need to merge the three lines to form two lines: 1) Pink + Purple 2) Green + Purple. The overlap should be fine.

Any help would be highly appreciated. Thanks!

Answer 1

• have used a few countries for source of LINESTRINGS
• with this data set, there are only two points where there are three selected countries that intersect (as per visualisation)
• key concept is inner join points to points, then find points that have more than required number of repetitions

import geopandas as gpd
import shapely.geometry
import pandas as pd
import plotly.express as px

# some polygons
# fmt: off
gdf = gpd.read_file(gpd.datasets.get_path("naturalearth_lowres")).loc[lambda d: d["iso_a3"].isin(["BEL", "LUX", "NLD", "DEU", "AUT","POL"]), ["geometry", "iso_a3"]].reset_index(drop=True)
# fmt: on

# change to linestrings as per question
gdf["geometry"] = gdf["geometry"].apply(
    lambda p: shapely.geometry.LineString(p.exterior.coords)
)

# generate series of points from linestrings
points = (
    gdf["geometry"].apply(lambda l: l.coords).rename("point").explode().reset_index()
)

# 1. join all points to all points
# 2. exclude point to self
points = (
    points.merge(points, on=["point"])
    .drop_duplicates()
    .loc[lambda d: d["index_x"] != d["index_y"]]
)
# now find points that appera N times
n_times = (
    points.groupby(["point"])
    .size()
    .loc[lambda s: s >= 3]
    .reset_index()["point"]
    .apply(pd.Series)
)

# visualize to test...
px.scatter_mapbox(n_times, lon=0, lat=1, mapbox_style="carto-positron",).update_traces(
    marker_size=20
).update_layout(mapbox={"layers": [{"source": gdf.__geo_interface__, "type": "line"}]})