基于特定规则对Shapely多边形列表进行分组的技术咨询

Got it, let's work through grouping your polys list of Shapely Polygons. Since you mentioned a "specific rule" but didn't dive into details, I'll walk through three common grouping scenarios you might be targeting, with ready-to-use code snippets tailored to your setup.

1. Group by Spatial Intersection/Touching #

If you want to group polygons that overlap, intersect, or touch each other (i.e., connected spatial components), here's a solid graph-based connected components approach:

from itertools import compress
from shapely.geometry import Polygon
import networkx as nx  # Install first via `pip install networkx`

# Your existing polygon list
polys = [<shapely.geometry.polygon.Polygon object at 0x000002D634217668>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217780>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D6341F9080>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217FD0>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217F60>]

# Build a graph where nodes = polygon indices, edges link intersecting polygons
G = nx.Graph()
G.add_nodes_from(range(len(polys)))

for i in range(len(polys)):
    for j in range(i+1, len(polys)):
        # Swap to .touches() if you only want polygons that touch (no overlap)
        if polys[i].intersects(polys[j]):
            G.add_edge(i, j)

# Extract connected components (each component is a group of polygon indices)
index_groups = list(nx.connected_components(G))

# Convert index groups back to actual polygon groups
polygon_groups = [[polys[idx] for idx in group] for group in index_groups]

# Check results
for idx, group in enumerate(polygon_groups):
    print(f"Group {idx+1}: {len(group)} connected polygons")

2. Group by Custom Attributes #

If your polygons have associated labels, area ranges, or other attributes, itertools.groupby is perfect. Let's assume you have a matching list of attributes for your polys:

from itertools import compress, groupby
from shapely.geometry import Polygon

# Your polygon list
polys = [<shapely.geometry.polygon.Polygon object at 0x000002D634217668>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217780>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D6341F9080>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217FD0>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217F60>]

# Example attribute list (matches order of polys)
poly_categories = ["residential", "commercial", "residential", "industrial", "commercial"]

# Sort pairs first (required for groupby to work correctly)
sorted_pairs = sorted(zip(polys, poly_categories), key=lambda x: x[1])

# Group polygons by their category
polygon_groups = {}
for category, group in groupby(sorted_pairs, key=lambda x: x[1]):
    polygon_groups[category] = [poly for poly, _ in group]

# Check results
for cat, group in polygon_groups.items():
    print(f"{cat.capitalize()} group: {len(group)} polygons")

For derived attributes like area ranges, replace the static category list with computed values:

# Group by area size (small/medium/large)
def get_area_bucket(poly):
    area = poly.area
    if area < 100:
        return "small"
    elif area < 500:
        return "medium"
    else:
        return "large"

sorted_pairs = sorted(zip(polys, [get_area_bucket(p) for p in polys]), key=lambda x: x[1])

3. Group by Spatial Clustering (Distance-Based) #

If you want to group polygons that are geographically close, use DBSCAN clustering on their centroid coordinates:

from itertools import compress
from shapely.geometry import Polygon
from sklearn.cluster import DBSCAN  # Install first via `pip install scikit-learn`
import numpy as np

# Your polygon list
polys = [<shapely.geometry.polygon.Polygon object at 0x000002D634217668>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217780>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D6341F9080>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217FD0>, 
         <shapely.geometry.polygon.Polygon object at 0x000002D634217F60>]

# Extract centroid coordinates for each polygon
centroids = np.array([[poly.centroid.x, poly.centroid.y] for poly in polys])

# Run DBSCAN (adjust eps and min_samples to fit your spatial scale)
dbscan = DBSCAN(eps=100, min_samples=1)  # eps = max distance between clustered points
cluster_labels = dbscan.fit_predict(centroids)

# Group polygons by their cluster label
polygon_groups = {}
for label in set(cluster_labels):
    # Filter polygons matching the current cluster label
    group = list(compress(polys, cluster_labels == label))
    polygon_groups[label] = group

# Check results
for label, group in polygon_groups.items():
    print(f"Cluster {label}: {len(group)} nearby polygons")

Just tweak the logic in whichever scenario fits your "specific rule"—if you had a different grouping criteria in mind, adjust the condition checks or clustering parameters accordingly!

内容的提问来源于stack exchange，提问作者Roman