在 geopandas 中使用缓冲区组合两个数据帧

Question

我有两个数据框（不是确切的数据，但相似）： df1：

Lon	Lat	Timestamp
4.44	61.41	2021-04-28 00:00:00
4.48	62.45	2021-04-28 00:02:00
4.51	61.48	2021-04-28 00:06:00
4.47	62.46	2021-04-28 00:08:00
4.44	61.41	2021-04-28 00:10:00
4.40	62.48	2021-04-28 00:12:00
4.51	61.44	2021-04-28 00:16:00
4.47	62.49	2021-04-28 00:18:00

df2

Lon	Lat	Timestamp
4.34	61.41	2021-04-28 00:00:00
4.38	62.45	2021-04-28 00:02:00
4.31	61.48	2021-04-28 00:06:00
4.17	62.46	2021-04-28 00:08:00
4.34	61.41	2021-04-28 00:10:00
4.30	62.48	2021-04-28 00:12:00
4.21	61.44	2021-04-28 00:16:00
4.47	62.49	2021-04-28 00:18:00

还有其他专栏，但我的问题与这些专栏有关。 所以我想为 df1 中的每个观察值在 100m 的半径内每分钟组合两个数据帧。

我只用一个数据帧做了类似的事情，对于数据帧中的每一次观察，我都加入了 100m 半径内的所有观察。

for name, group in df.groupby(['timestamp']):
        buf = group.copy()
        buf['geometry'] = buf.geometry.buffer(100)
        points_within = gpd.sjoin(group, buf,   op = 'within')

我需要用两个数据框做类似的事情

Answer 1

• 样本数据集中 100m 范围内的内容不多。增加距离意味着更多sjoin()
• 将 GeoPandas 功能与 CRS 和 buffer() 结合使用。重要的是 UTM 几何用于距离。因此投影到 UTM 并返回到 EPSG:4326
• 已显示输出数据帧和 plotly mapbox 作为标记和包含缓冲区的 geojson

import geopandas as gpd
import shapely, json
import pandas as pd
import plotly.express as px

df1 = pd.DataFrame(
    {
        "Lon": [4.44, 4.48, 4.51, 4.47, 4.44, 4.4, 4.51, 4.47],
        "Lat": [61.41, 62.45, 61.48, 62.46, 61.41, 62.48, 61.44, 62.49],
        "Timestamp": [
            "2021-04-28 00:00:00",
            "2021-04-28 00:02:00",
            "2021-04-28 00:06:00",
            "2021-04-28 00:08:00",
            "2021-04-28 00:10:00",
            "2021-04-28 00:12:00",
            "2021-04-28 00:16:00",
            "2021-04-28 00:18:00",
        ],
    }
)

df2 = pd.DataFrame(
    {
        "Lon": [4.34, 4.38, 4.31, 4.17, 4.34, 4.3, 4.21, 4.47],
        "Lat": [61.41, 62.45, 61.48, 62.46, 61.41, 62.48, 61.44, 62.49],
        "Timestamp": [
            "2021-04-28 00:00:00",
            "2021-04-28 00:02:00",
            "2021-04-28 00:06:00",
            "2021-04-28 00:08:00",
            "2021-04-28 00:10:00",
            "2021-04-28 00:12:00",
            "2021-04-28 00:16:00",
            "2021-04-28 00:18:00",
        ],
    }
)

MIN_DIST = 10**2

gdf1 = gpd.GeoDataFrame(
    geometry=df1.loc[:, ["Lon", "Lat"]]
    .apply(lambda r: shapely.geometry.Point(r["Lon"], r["Lat"]), axis=1)
    .values,
    crs="EPSG:4326",
)

gdf2 = gpd.GeoDataFrame(
    geometry=df2.loc[:, ["Lon", "Lat"]]
    .apply(lambda r: shapely.geometry.Point(r["Lon"], r["Lat"]), axis=1)
    .values,
    crs="EPSG:4326",
)

# add buffer to df1,  NB need to correctly use CRS systems to define distances
gdf1 = (
    gdf1.to_crs(gdf1.estimate_utm_crs()).geometry.buffer(MIN_DIST).to_crs("EPSG:4326")
)

# join data frames back together
df2_in_df1 = df2.reset_index().merge(
    gpd.sjoin(gpd.GeoDataFrame(geometry=gdf1), gdf2, how="inner"),
    left_on="index",
    right_on="index_right",
)


# plot it to see what's been found
fig = (
    px.scatter_mapbox(df1, lat="Lat", lon="Lon")
    .update_traces(marker={"color": "red", "opacity":.3})
    .add_traces(px.scatter_mapbox(df2, lat="Lat", lon="Lon").update_traces(marker={"color":"red", "opacity":.3}).data)
    .add_traces(px.scatter_mapbox(df2_in_df1, lat="Lat", lon="Lon").update_traces(marker={"color":"green", "size":10}).data)

    )

fig.update_layout(
    mapbox={
        "style": "open-street-map",
        "layers": [
            {
                "source": json.loads(gdf1.geometry.to_json()),
                "below": "traces",
                "type": "line",
                "color": "purple",
                "line": {"width": 1.5},
            }
        ],
    },
    margin={"l": 0, "r": 0, "t": 0, "b": 0},
)

	index	Lon	Lat	Timestamp	index_right
0	7	4.47	62.49	2021-04-28 00:18:00	7