OSMnx：一组起点和终点的最短路径

Question

假设我有一个包含起点和终点的点 shapefile。

1) 如何使用 nx.shortest_path 计算每个起点和终点？ 2）如何将对应的路由保存为shapefile？我检查了Save a route and conserve its curvature with Python OSMnx，它显示了如何为路线获取 MultiLineString，但没有显示如何导出路线。

Answer 1

以下步骤可行：

1. 用 geopandas 打开你的 shapefile
2. 对于 shapefile 中的每个原点，使用 OSMnx 找到 nearest network node，然后对每个目标点执行相同操作
3. 使用 nx.shortest_path 计算每个起点/终点节点对之间的最短路径
4. 保存路线to a MultiLineStringwkt
5. 由于 wkt 是纯文本，只需使用任何 Python 序列化方法将此文本保存到磁盘即可。或者将所有 MultiLineStrings 本身组装成 geopandas GeoSeries 并将其作为 shapefile 或 GeoJSON 文件保存到磁盘。

Answer 2

感谢@gboeing 提供的步骤。还添加了对我有用的代码 sn-ps 来实现概述的解决方案。

from shapely.geometry import shape
from shapely.geometry import LineString

from osgeo import ogr, osr
import geopandas as gpd
import pandas as pd
import time
import networkx as nx
import osmnx as ox


G= ox.graph_from_place('Bangalore, India')

fig, ax = ox.plot_graph(G)

#Read the Origin-Destination csv
od_table = 'E:/OD_pairs1.csv'
df = pd.read_csv(od_table)

以下是我的输入 CSV 与原点 (o_long, o_lat) 和目标 (d_long, d_lat) 点坐标的样子：

df

Out[]:  o_id  o_long    o_lat      d_id d_long      d_lat
        o1  77.548349   12.996800   d1  77.554137   12.995225
        o2  77.555820   13.009082   d2  77.570458   12.995690
        o3  77.576325   13.014630   d3  77.583616   13.009188
        o4  77.564848   12.990121   d4  77.551316   12.988570
        o5  77.590529   12.992340   d5  77.598469   12.988289

答案的第 1 部分（计算路线的方法）： （在 shortestpath() 方法中，如果您不想将 O&D 点到最近网络节点的距离添加到路由长度中，您可能希望返回 length 变量而不是 total_length。）

def nodes_to_linestring(path):
    coords_list = [(G.nodes[i]['x'], G.nodes[i]['y']) for i in path ]
    #print(coords_list)
    line = LineString(coords_list)
    
    return(line)

def shortestpath(o_lat, o_long, d_lat, d_long):
    
    nearestnode_origin, dist_o_to_onode = ox.distance.get_nearest_node(G, (o_lat, o_long), method='haversine', return_dist=True)
    nearestnode_dest, dist_d_to_dnode = ox.distance.get_nearest_node(G, (d_lat, d_long), method='haversine', return_dist=True)
    
    #Add up distance to nodes from both o and d ends. This is the distance that's not covered by the network
    dist_to_network = dist_o_to_onode + dist_d_to_dnode
    
    shortest_p = nx.shortest_path(G,nearestnode_origin, nearestnode_dest) 
    
    route = nodes_to_linestring(shortest_p) #Method defined above
    
    # Calculating length of the route requires projection into UTM system.  
    inSpatialRef = osr.SpatialReference()
    inSpatialRef.ImportFromEPSG(4326)
    outSpatialRef = osr.SpatialReference()
    outSpatialRef.ImportFromEPSG(32643)
    coordTransform = osr.CoordinateTransformation(inSpatialRef, outSpatialRef)
    
    #route.wkt returns wkt of the shapely object. This step was necessary as transformation can be applied 
    #only on an ogr object. Used EPSG 32643 as Bangalore is in 43N UTM grid zone.
    geom = ogr.CreateGeometryFromWkt(route.wkt)
   
    geom.Transform(coordTransform)
    length = geom.Length()
    
    #Total length to be covered is length along network between the nodes plus the distance from the O,D points to their nearest nodes
    total_length = length + dist_to_network
    #in metres
    
    return(route, total_length )

在数据帧上应用上述方法来获取所有 O-D 对的最短路径的几何形状和长度：

start_time = time.time()

df['osmnx_geometry'] = df.apply(lambda x: shortestpath(x['o_lat'], x['o_long'], x['d_lat'], x['d_long'])[0] , axis=1)
df['osmnx_length'] = df.apply(lambda x: shortestpath(x['o_lat'], x['o_long'], x['d_lat'], x['d_long'])[1] , axis=1)

print("Time taken: ", (time.time() - start_time), "seconds")

生成的数据框将具有路线的相关几何形状和路线长度（以米为单位）：

df
    Out[]:
    o_id    o_long  o_lat   d_id    d_long  d_lat      osmnx_geometry     osmnx_length
    o1  77.548349   12.996800   d1  77.554137   12.995225   LINESTRING (77.5482836 12.9966618, 77.54976259...   827.718256
    o2  77.555820   13.009082   d2  77.570458   12.995690   LINESTRING (77.555814 13.0090627, 77.5556026 1...   2588.006507
    o3  77.576325   13.014630   d3  77.583616   13.009188   LINESTRING (77.57588320000001 13.0146859, 77.5...   1107.137060
    o4  77.564848   12.990121   d4  77.551316   12.988570   LINESTRING (77.56473080000001 12.9898858, 77.5...   1744.708360
    o5  77.590529   12.992340   d5  77.598469   12.988289   LINESTRING (77.5901456 12.9920295, 77.5905355 ...   1097.493520

答案的第 2 部分。将此数据框保存为 shapefile：

df = df.rename(columns = {'osmnx_geometry': 'geometry'})
gpdf = gpd.GeoDataFrame(df, geometry =df['geometry'])
gpdf.to_file('osmnx_shortestpaths.shp')

我还探索了 OSRM（用于最快路线）和 Gmaps Directions API（用于基于流量的最有效路线）来计算同一组 O-D 对的路线，可以在 here 找到相同的脚本。