-
Notifications
You must be signed in to change notification settings - Fork 80
user manual
The osm-fieldwork project is a collection of utilities useful for field data collection, focused on OpenStreetMap (OSM) and ODK. Both of these are used heavily for humanitarian and emergency response by many organizations. The problem is these two projects were never designed to work together, so this project was born to fill the gaps. Included are a few other useful utilities for field mapping.
This project is also currently part of the backend for the FMTM project, but all of the data processing can also be run standalone, and also works fully offline. All the standalone programs run in a terminal, and are written in Python.
ODK is a format for collecting data on mobile devices, including the spatial coordinates of that data item. The primary source file is a spreadsheet called an XLSForm, which gets converted to an XForm using the xls2xform program. An XForm is is in XML format. All collected data is stored as an instance file, also in XML format on the mobile device, but of course is a different schema than the XForm. Once the data is collected it gets uploaded to an ODK Central server. From there you can download the collected data, called submissions, in CSV or JSON format. The JSON format works better. Here's where the conversion project starts, how to process the downloaded data into something we can upload to OpenStreetMap efficiently.
All of the XLSForms included in this project have all been carefully edited to enable a good clean conversion to OSM XML. More information on how to modify the conversion is here. If you base any custom XLSForms from this library, you can also update the conversion criteria. These XLSForms can also be downloaded from FMTM.
The FMTM is a project to coordinate field data collection in a similar way as the HOT Tasking Manager. But other than the ability to break up a big area into tasks, the rest works very differently. Often mangaing a group doing field mapping is a bit like herding cats. Plus the mappers often aren't sure where they should be mapping, or when they are finished. In addition, it is now possible to load a data extract from OSM into ODK Collect, and use that data to set the default values when collecting the data so the mapper doesn't have to do it. FMTM handles the creation of the data extract, as well as processing the data into a format suitable to edit with JOSM or QGIS. The FMTM backend is a FastAPI wrapped around this project.
This project is available from PyPi.org, and can be installed like this:
pip install osm-fieldwork
It contains multiple programs, each one that handles a specific part of the conversion process. Each program is a single class so it can be used as part of a FastAPI backend, but also runs standalone for debugging, and working offline. These are all terminal based, as the website frontend is the actual GUI.
- json2osm
- Convert JSON from Central to OSM XML
- csv2osm
- Convert CSV from Central to OSM XML
- odk2csv
- Convert the ODK Instance to CSV
- odk2geojson
- Convert the ODK Instance to GeoJson
- parsers
- Conflate POIs from Collect with existing OSM data
- odk_client
- Remotely control an ODK Central server
You can also to run the terminal based programs from the source tree, which can be gotten from here:
git clone [email protected]:hotosm/osm-fieldwork.git
This section will focus on converting the JSON format, but the process for converting the CSV submissions is the same. The JSON format seems to be more complete for some XLSForms, so it's preferred. The first step is converting it to OSM XML format, so it can be loaded into JOSM and edited. A YAML based config file is used to convert the JSON format you just downloaded into the OSM XML format.
The initial problem is neither the CSV or the JSON format stores the coordinates in a way any editing program wants them. So that's the most important part of the conversion process, generating a data file with spatial coordinates in the right syntax. The conversion process generates two output files, one in OSM XML format, the other in GeoJson format. The OSM XML one has had the data filtered, not everything collected is for OSM. But all the data goes in the GeoJson file, so nothing is lost. Since the GeoJson format does not have to follow OSM syntax, not all the tags and values may be similar to what OSM expects, but that's not a problem for our use case.
The config file for conversion has 3 sections, one for all the conversion data, one for data to ignore completely, and a private section for the GeoJson file. The stuff to ignore is extraneous fields added by ODK Collect, like deviceID. Modifying the conversion is straight forward as it's mostly just replacing one set of strings with another.
For any of the XLSForms in this project's library, the configuration is already done, but any custom XLSForms will need to modify it to get a good conversion, or fix it in JOSM later. For a one-off project, like an import, I usually get lazy and fix it in JOSM. But for anything used several times, that gets old, so it's better to improve the config file.
To convert the JSON format file downloaded for ODK Central, run this program:
json2osm.py -i Submissions.json json2osm.py -i Submissions.json -y custom.yaml
or for the CSVfile:
CSVDump.py -i Submissions.csv
CSVDump.py -i Submissions.csv -u custom.yaml
which produces a Submissions.osm and Submissions.geojson files from that data. The OSM XML file may have tags that got missed by the conversion process, but the advantage is now all the data can be viewed and edited by JOSM. If you want a clean conversion, edit the config file and use that as an alternate for converting the data.
json2osm -i Submissions.json -x custom.yaml
Now you have a file that can be viewed or edited, but it's all collected, but some of the features may already exist in OSM. This can be done manually in JOSM, which is ok for small datasets, but it's easier to apply a little automated help. It's possible to find similar features in OSM that are near the data we just collected for a building, but has the same business name. How to conflate the collected data with existing OSM data is another document.
To just use the conflation software requires setting up a postgres database containing the OSM data for the county, region, state, country, ect... You can also use the data extract from FMTM, as it covers the same area the data was collected in. FMTM allow you to download the data extract used for this task. Postgres works much faster, but the GeoJson data extract works too as the files per task are relativly small.
odk_merge.py Submissions.osm PG:"nepal" -b kathmanu.geojson
or
odk_merge.py Submissions.osm kathmandu.geojson
In this example, the OSM XML file from the conversion process uses a local postgres data with the country of Nepal loaded into it. You can also specify an alternate boundary so the conflation will use a subset of the entire database to limit the amount of data that has to be queried.
Each feature in the submission is queried to find any other features with 2 meters where any tags match. Both POIs and buildings are checked for a possible match. Often the building has "building=yes" from remote mapping, so we'd also want to merge the tags from the collected data into the building way. Multiple shops within the same building remain as a POI in that building.
There is much more detail on this program here.
Basemaps are very useful when using ODK Collect in areas where the map data is poor. Imagery is particular is very useful, as you can use that to select a location other than where you are standing. This project has a utility that makes basemaps from several sources. It builds a local tile store, so larger areas can be downloaded and in the field when offline, smaller basemaps can be made from the tile store. Since downloading map tiles is very time consuming, I usually download larger areas and let it download for a few days.
basemapper -s esri -b Pokara.geojson -z 8-15 -o pokara.mbtiles
This command will download all the map tiles from ESRI into an XYZ tile store for zoom levels 8 to 15. Since downloading imagery is slow, I often download larger areas, and then use a subset of the tiles to make smaller basemaps. The mbtiles file can be manually loaded into ODK Collect as a layer, and used to adjust the location of the POI when mapping.
Since it often useful for navigation, basemapper can also produce a basemap from the same map tiles for Osmand. This is very useful when in areas with little map data, for example during a remote backcountry trip. This example downloads Bing imagery for Pokara, Nepal.
basemapper -s bing -b Pokara.geojson -z 8-19 -o pokara.sqlitedb
There is much more detail on this program here.
These programs read the XML format used by ODK Collect for Instance files. Since each submission has a separate Instance file, this takes a regular expression, and produces a single output file. This is only used when working offline, so it's possible to edit the recently collected data and update the map data. Very useful when working offline during big disasters.
odk2osm -i Highways Paths_2023-07-17\*
On your phone, you can find the instance files here:
/sdcard/Android/data/org.odk.collect.android/files/projects/[UUID]/instances
You can also manually update your data extracts by copying them to /sdcard/Android/data/org.odk.collect.android/files/projects/[UUID]/forms/[Form name]-media/
And manually update the XForm by copying them to /sdcard/Android/data/org.odk.collect.android/files/projects/[UUID]/forms/
[ODK Central](https://docs.getodk.org/central-intro/ is the server side of ODK Collect. It's where XForms are downloaded from, and where submissions go after being sent by Collect. As there are a lot of options, this program is not very user friendly as it's primarily used as part of the backend for the FMTM project, and most people would just use the Central website.
However, this can be useful for scripting the server. For example to list all the projects on a remote Central server:
odk_client -s projects
And this lets you download all the submissions to project number 19 and using the XLSForm for buildings.
odk_client -v -i 19 -f buildings -x json
There is much more detail on this program here.