GeoPandas Spatial Join – How to Spatially Join Only Features by Largest Overlap with Sjoin in GeoPandas

Question

I am using GeoPandas to join two GeoDataFrames with the parameter how="inner", op="intersects".

Due to nature of the files a larger percentage of the files are duplicates (in regard of the geometry). Therefore I want to ask how to filter out and restrict to a single area, like selecting only output file entries with the highest Intersection over Union, DICE coefficient, Overlap coefficient or similar of the overlapping geometries.

Answer 1

Intersect, sort by area and drop duplicates:

import geopandas as gpd
import psycopg2

con = psycopg2.connect(database="lmv", user="postgres", password="dsfdas",
    host="localhost")

df1 = gpd.GeoDataFrame.from_postgis("select geom, lan_kod from ok_an_riks", con, geom_col='geom' )
df2 = gpd.GeoDataFrame.from_postgis("select geom, ogc_fid from ok_nd_riks", con, geom_col='geom' )

df3 = gpd.overlay(df1, df2, how='intersection')

#Sort by area so largest area is last
df3.sort_values(by=df3.geometry.area, inplace=True)

#Drop duplicates, keep last/largest
df3.drop_duplicates(subset='ogc_fid', keep='last', inplace=True)

Example:

print(df3.loc[df3['ogc_fid']==105][['lan_kod','ogc_fid']])
   lan_kod  ogc_fid
42      04      105