如何从OSMnx图中筛选指定边并构建子图？

问题背景 #

• 下载OSM地图为shapefile格式，筛选出目标路线的节点与特定街道边，导出为CSV后导入Python Notebook
• 通过G.subgraph(nodes)成功生成仅包含目标节点的子图，但无法筛选目标边
• 尝试使用G.edge_subgraph(edges)，分别传入OSM ID列表[osmid_0,...,osmid_i]和节点对列表[(u_0,v_0),...(u_i,v_i)]，执行G_mh.edge_subgraph(roads['osmid'])时触发以下错误：

Output exceeds the size limit. Open the full output data in a text editor---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[230], line 1
----> 1 G_mh.edge_subgraph(roads['osmid'])

File ~/Library/Python/3.9/lib/python/site-packages/networkx/classes/graph.py:1865, in Graph.edge_subgraph(self, edges)
   1827 def edge_subgraph(self, edges):
   1828     """Returns the subgraph induced by the specified edges.
   1829 
   1830     The induced subgraph contains each edge in `edges` and each
   (...)
   1863 
   1864     """
-> 1865     return nx.edge_subgraph(self, edges)

File ~/Library/Python/3.9/lib/python/site-packages/networkx/classes/function.py:445, in edge_subgraph(G, edges)
    443 if G.is_multigraph():
    444     if G.is_directed():
-> 445         induced_edges = nxf.show_multidiedges(edges)
    446     else:
    447         induced_edges = nxf.show_multiedges(edges)

File ~/Library/Python/3.9/lib/python/site-packages/networkx/classes/filters.py:69, in show_multidiedges(edges)
     68 def show_multidiedges(edges):
---> 69     edges = {(u, v, k) for u, v, k in edges}
...
     68 def show_multidiedges(edges):
---> 69     edges = {(u, v, k) for u, v, k in edges}
     70     return lambda u, v, k: (u, v, k) in edges

ValueError: too many values to unpack (expected 3)

解决方案 #

错误原因

OSMnx默认生成的是MultiDiGraph（多有向图），edge_subgraph对这类图要求传入**(u, v, key)**三元组作为边的唯一标识符，而不是单独的osmid或(u,v)二元组。另外，osmid是边的属性字段，不是图的边标识符，直接传入osmid列表会触发解包错误。

具体实现步骤

1. 通过osmid筛选边并生成子图
   从CSV中提取目标osmid，遍历原图匹配对应边的三元组：

   import osmnx as ox
   import pandas as pd
   
   # 假设roads是导入的CSV数据框，包含目标边的osmid列
   target_osmids = set(roads['osmid'])
   
   # 遍历原图的所有边，收集符合条件的(u, v, key)三元组
   selected_edges = []
   for u, v, key, edge_data in G_mh.edges(data=True, keys=True):
       # 注意部分边的osmid可能是列表（如合并的路段），需特殊处理
       edge_osmid = edge_data.get('osmid')
       if isinstance(edge_osmid, list):
           if any(oid in target_osmids for oid in edge_osmid):
               selected_edges.append((u, v, key))
       else:
           if edge_osmid in target_osmids:
               selected_edges.append((u, v, key))
   
   # 生成边诱导子图
   G_sub = G_mh.edge_subgraph(selected_edges)
   

2. 结合节点ID与osmid精准匹配
   如果CSV中包含图的节点ID（u和v列），可以更精准地定位边：

   selected_edges = []
   for _, row in roads.iterrows():
       u, v, target_oid = row['u'], row['v'], row['osmid']
       # 遍历该(u,v)对应的所有key（多图特性）
       if u in G_mh and v in G_mh[u]:
           for key in G_mh[u][v]:
               edge_data = G_mh[u][v][key]
               edge_osmid = edge_data.get('osmid')
               # 处理osmid为列表的情况
               match = False
               if isinstance(edge_osmid, list):
                   match = target_oid in edge_osmid
               else:
                   match = (edge_osmid == target_oid)
               if match:
                   selected_edges.append((u, v, key))
                   break  # 找到匹配项后退出循环
   G_sub = G_mh.edge_subgraph(selected_edges)
   

3. 保存子图
   生成子图后，可通过以下方式保存：

   # 保存为GraphML格式（保留所有属性，推荐）
   ox.save_graphml(G_sub, filepath='target_subgraph.graphml')
   
   # 保存为shapefile（适合GIS工具处理）
   ox.save_graph_shapefile(G_sub, filepath='target_subgraph')
   

内容的提问来源于stack exchange，提问作者sdb