map-matching of GPS traces, built using pgrouting

For details of the algorithm, see Millard-Ball, Adam; Hampshire, Robert and Weinberger, Rachel (2019), "Map-matching poor-quality GPS data in urban environments: The pgMapMatch package." Transportation Planning and Technology. (A non-paywalled pre-print is available here.)

For questions and feedback, please contact Adam Millard-Ball.

1. Download this repository using git clone https://github.com/amillb/pgMapMatch.git

2. You will need to load a table of street edges into a PostgreSQL database. The easiest way to do this is:

• Download an extract from OpenStreetMap, such as those produced for metropolitan areas by Mapzen.

• Use osm2po to load the OpenStreetMap data into PostgreSQL.

• Transform the street geometries to a suitable projection, for example:

  ALTER TABLE streets_table ALTER COLUMN geom_way TYPE Geometry(LineString, your_srid) USING ST_Transform(geom_way, your_srid);

3. Copy config_template.py to config.py, and adjust the configuration settings. You will almost certainly need to adjust the postgres login information in the pgInfo dictionary. If you don't use osm2po, you may also need to adjust the column names for the streets table specified in config.py. The default parameters in config_template.py assume your units are in meters, so it will be easiest to use a projection that is also in meters. If you use feet, you will need to change the parameters, and/or risk getting unexpected results.

You can match traces from GPX files (a time field must be included), or a PostgreSQL table. For the postgres option, the traces must be LineStrings with an M coordinate providing the timestamp of each point, and the projection must be the same as the table of streets.

You can call pgMapMatch from the command line. You can match either a GPX file, or a table of GPS traces loaded into PostgreSQL. python pgMapMatch --help gives you a list of options.

You can also import pgMapMatch into Python, and use the class mapMatcher(). After you call matchGPXTrace() or matchPostgresTrace(), you can access the sequence of edges, matched geometry and match score, and write them to Postgres. For example:

import pgMapMatch
mm = pgMapMatch.mapMatcher('streetsTable')
mm.matchGPXTrace(gpx_filename)
mm.bestRoute        # returns the sequence of edge ids (based on the id column in the streets table)
mm.getMatchAsWKT()  # returns the matched geometry as Well-Known Text
mm.getMatchScore()  # returns the match score (probability that the match is good)

PostgreSQL, with PostGIS 2.3+ and pgrouting 2.4.1+ installed. A local installation is not required; PostgreSQL can run on a remote server. Make sure to update pgInfo in config.py with the database connection information.

The following Python packages:

• numpy 1.11.3+
• scipy 0.19.0+
• pandas 0.19.2+
• gpxpy 1.1.2+
• psycopg2 2.5.2+
• sqlalchemy 1.1.6+
• docopt 0.6.1+

Other versions may work, but have not been tested.

You can install all the Python packages with: pip install numpy scipy pandas gpxpy psycopg2 sqlalchemy docopt