使用OSMnx的project_graph方法报错：'DataFrame'无'crs'属性

运行代码 #

# 从地址获取路网，仅包含地址沿路网1km范围内的节点
G = ox.graph_from_address(
    address="350 5th Ave, New York, NY",
    dist=1000,
    dist_type="network",
    network_type="drive",
)

# 将路网投影到UTM（自动计算投影带）
G_projected = ox.project_graph(G)

报错信息 #

AttributeError                            Traceback (most recent call last)
Cell In [22], line 10
      2 G = ox.graph_from_address(
      3     address="350 5th Ave, New York, NY",
      4     dist=1000,
      5     dist_type="network",
      6     network_type="drive",
      7 )
      9 # you can project the network to UTM (zone calculated automatically)
---> 10 G_projected = ox.project_graph(G)

File ~/opt/anaconda3/envs/ox/lib/python3.11/site-packages/osmnx/projection.py:144, in project_graph(G, to_crs)
    141 if "simplified" in G.graph and G.graph["simplified"]:
    142     # if graph has previously been simplified, project the edge geometries
    143     gdf_edges = utils_graph.graph_to_gdfs(G, nodes=False, fill_edge_geometry=True)
--> 144     gdf_edges_proj = project_gdf(gdf_edges, to_crs=gdf_nodes_proj.crs)
    145 else:
    146     # if not, you don't have to project these edges because the nodes
    147     # contain all the spatial data in the graph (unsimplified edges have
    148     # no geometry attributes)
    149     gdf_edges_proj = utils_graph.graph_to_gdfs(G, nodes=False, fill_edge_geometry=False).drop(
    150         columns=["geometry"]
    151     )
...
   5900 ):
   5901     return self[name]
-> 5902 return object.__getattribute__(self, name)

AttributeError: 'DataFrame' object has no attribute 'crs'

排查结果 #

已确认 G.graph['crs'] 返回值为 'epsg:4326'

问题根源 #

通过conda安装的OSMnx版本为1.0.1，与示例文件依赖的1.2.2版本不符，版本差异导致该错误。

解决步骤 #

通过pip安装指定版本的OSMnx，操作如下：

conda create -n ox_that_work
conda activate ox_that_work
conda install pip
# 安装OSMnx依赖包
pip install -r requirements.txt
# 安装1.2.2版本的OSMnx
pip install osmnx==1.2.2
# 安装依赖列表中未包含的descartes和scikit-learn
pip install descartes scikit-learn

内容的提问来源于stack exchange，提问作者lorenzo