This python script allows map matching (matching of tracking points to a network) in arcpy using a Hidden Markov model with probabilities parameterized based on spatial + network distances. Follows the ideas in Newson, Krumm (2009): "Hidden markov Map Matching through noise and sparseness"
Author: Simon Scheider
Created: 16/03/2017
To install as a toolbox in ArcGIS, see mapMaptch.pyt
The code is written in Python 2.7 and depends on:
- arcpy (ships with ArcGIS and its own Python 2.7)
- networkx
python pip install networkx
- note: requires installing GDAL first, which can be obtained as a windows wheel from here and then installed with pip locally:
python pip install GDAL-2.1.3-cp27-cp27m-win32.whl
To install the mapmatcher Python module, simply download and execute this windows executable:
Example:
from mapmatcher import mapmatcher
arcpy.env.workspace = 'C:/Users/simon/Documents/GitHub/mapmatching'
opt = mapmatcher.mapMatch('testTrack.shp', 'testSegments.shp')
#outputs testTrack_pth.shp
mapmatcher.exportPath(opt, 'testTrack.shp')
The last method saves a new shape file named testTrack_pth.shp in the current arcpy workspace, containing a sequence of segments to which the track was mapped.
Results are shown here:
The main method is mapMatch. Based on the Viterbi algorithm for Hidden Markov models, see https://en.wikipedia.org/wiki/Viterbi_algorithm, it gets trackpoints and segments, and returns the most probable segment path (a list of segments) for the list of points.
-
@param track = a shape file (filename) with point geometries representing a track, can be unprojected (WGS84). The order of points in this file should reflect the temporal order.
-
@param segments = a shape file of network segments, should be projected (in meter) to compute Euclidean distances properly (e.g. GCS Amersfoord). Note: To compute network distances, the script turns this network into a graph using networkx, based on coincidence of segment end points. It is therefore important that logically connected segments are also geometrically connected (no geometrical errors). Other than this, the script does have other requirements for the network.
-
@param decayconstantNet (optional) = the network distance (in meter) after which the match probability falls under 0.34 (exponential decay). Default is 30 meters. This distance parameter depends on the intervals between successing points in the track.
-
@param decayConstantEu (optional) = the Euclidean distance (in meter) after which the match probability falls under 0.34 (exponential decay). Default is 10 meters. This distance parameter depends on the measurement accuracy of tracking points.
-
@param maxDist (optional) = the Euclidean distance threshold (in meter) for taking into account segments candidates. Default is 50 meters. Depends also on measurement accuracy of track points.
-
result = delivers back a path (a list of segment ids).
Depending on the type of movement, optional parameters need to be fine tuned to get optimal results. For example, when tracking frequency is very slow, then track points are far apart, and then decayconstantNet needs to be increased accordingly.
Exports the path into a shape file named segments_pth.shp inside the current ArcGIS workspace.
To use the Python method as an ArcGIS toolbox, you need to do the following:
- In your ArcGIS Python version (e.g. Folder
C:\Python27\ArcGIS10.3\Lib\site-packages), install required modules for GDAL and networkx in a cmd window:
- if you have not installed it yet, install pip (http://pip.readthedocs.io/en/latest/installing/).
- Download a suitable GDAL wheel from here. Then execute:
python pip install GDAL-2.1.3-cp27-cp27m-win32.whlpython pip install networkx
-
Install mapmatcher Python module by downloading and executing the windows executable mapmatching/dist/mapmatcher-1.0.win32.exe. Make sure you select exactly the Python installation that ships with your ArcGIS as a target folder.
-
Download the ArcGIS Python toolbox mapMatch.pyt, together with meta data files mapMatch.mapMatch.pyt.xml and mapMatch.pyt.xml and drop it anywhere on your computer.
-
Now you can open the toolbox by clicking on it inside an ArcGIS Catalog Window:
The tool saves a new shape file named NameofInputTrack_pth.shp inside the current ArcGIS workspace that contains the path of segments to which the track was mapped. When executing, make sure the network is as small as possible to speed up.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent