GeoPandas and OpenStreetMap¶

The GeoPandas osm module makes it easy to dynamically query and analyze arbitrary OpenStreetMap (OSM) data. It's a powerful combination.

This example finds the number of traffic lights that are along an arbitrary running route, using just a few lines of code. Note that OpenStreetMap doesn't have lots of traffic lights, but they are pretty good around Cambridge, MA where this example is located.

I used RunKeeper to download a recorded run from my phone. You can find the raw file here:

https://gist.githubusercontent.com/jwass/4ae78872e30a21b34a44/raw/e48b18797e2dd9fb653d38329df574fed1852057/RK_gpx%20_2014-07-10_2054.gpx

I removed a bunch of random points near the start and end so you can't exactly figure out where I live :)

import numpy as np
import geopandas as gpd
import geopandas.io.osm as osm

Load the trace¶

First load up the GPX file. Specifying the tracks layer loads the data as a single row containing a LineString or MultiLineString. For other uses you can specify layer=track_points to have a row for each recorded point.

df = gpd.read_file('RK_gpx_2014-07-10_2054.gpx', layer='tracks')
df

Load the OpenStreetMap data¶

Now load all the OpenStreetMap traffic lights in the route's bounding box. Traffic lights in OSM are represented as nodes where the highway tag is traffic_signals. The query_osm function takes care of formulating the query and returning a GeoDataFrame ready to go with all the points.

df_lights = osm.query_osm('node', bbox=df.total_bounds, tags='highway=traffic_signals')
df_lights

Project both GeoDataFrames to the Massachusetts state plane (EPSG:26986) (Hopefully in the future GeoPandas will have geography methods and we can skip the projections).

Then just compute the distance from the lights to the line using distance().

epsg = 26986  # MA mainland state plane
df_p = df.to_crs(epsg=epsg)
df_lights_p = df_lights.to_crs(epsg=epsg)

df_lights['d'] = df_lights_p.distance(df_p['geometry'].iloc[0])
df_lights['d']

0      13.632391
1       2.799502
2     382.609745
3      11.127261
4     150.686201
5     308.685223
6     129.537163
7     259.754943
8      17.562312
9      30.430179
10    555.348314
11     86.198503
12    206.003460
13    182.843900
14      5.179193
15     90.214907
16     83.550469
17    147.394822
18     12.741685
Name: d, dtype: float64

Visualization¶

Now visualize the data. I'm using geojsonio.py which opens http://geojsonio.io. We can take advantage of simplestyle-spec by setting a few properties on the GeoDataFrame to control plotting behavior

Use the marker-color property - traffic lights that are close to the route (< 20 meters) are green and the others will be red. The light at Massachusetts Ave. & Landsdowne St. is a false negative due to noise in the GPS trace.

You can click on each marker and inspect the d property to see how far it is from the route.

# Shouldn't need the 'set_geometry()' call below, this is a problem in GeoPandas
combined = df.append(df_lights).set_geometry('geometry')
combined['marker-color'] = np.where(combined['d'] <= 20, '#4daf4a', '#e41a1c')

import geojsonio
geojsonio.embed(combined.to_json(na='drop'))

	highway	id	traffic_signals	geometry
0	traffic_signals	61283269	NaN	POINT (-71.09488469999999 42.3601506)
1	traffic_signals	61283293	NaN	POINT (-71.0965425 42.3561028)
2	traffic_signals	61317400	NaN	POINT (-71.1135493 42.3622591)
3	traffic_signals	61318066	NaN	POINT (-71.1046257 42.3535672)
4	traffic_signals	61321123	blinker	POINT (-71.11036199999999 42.357534)
5	traffic_signals	61321134	blinker	POINT (-71.10724089999999 42.3581054)
6	traffic_signals	61321277	NaN	POINT (-71.09943370000001 42.3634087)
7	traffic_signals	61322052	NaN	POINT (-71.110533 42.363275)
8	traffic_signals	61323022	NaN	POINT (-71.0960005 42.3608275)
9	traffic_signals	61323032	NaN	POINT (-71.0975461 42.361748)
10	traffic_signals	61327037	NaN	POINT (-71.113873 42.3644915)
11	traffic_signals	61327067	NaN	POINT (-71.101387 42.364001)
12	traffic_signals	61327121	NaN	POINT (-71.0967089 42.3632016)
13	traffic_signals	61328029	blinker	POINT (-71.1127706 42.3601424)
14	traffic_signals	61328070	NaN	POINT (-71.11107869999999 42.3591161)
15	traffic_signals	61329632	NaN	POINT (-71.10118799999999 42.363889)
16	traffic_signals	61329844	NaN	POINT (-71.099593 42.362982)
17	traffic_signals	61331756	NaN	POINT (-71.1095894 42.3582726)
18	traffic_signals	579060230	NaN	POINT (-71.0935624 42.3590304)