OpenStreetMap
There are multiple mapping tasks in OpenStreetMap that people still have to do old-style, like we did in 2010. But with current technology and with current funding of OSM research, these can be automated, improving mapping quality over the world.
If you work on OSM-related projects for a company, like in Facebook or Apple, please ask your manager if they can spare your time on projects useful for the community. For example:
-
Imagery alignment! People from the US and Western Europe won’t understand, but in most parts of the world, imagery is offset up to hundreds of meters. Mappers need to shift it to start mapping, otherwise different people would map with different offsets.
We have plugins for automatically tracing roads, buildings and rivers. Why not create a button for JOSM and iD that looks at GPS traces (rasterized or not) and finds the best offset for an imagery layer?
-
okay I had just that one idea and struggle to come up with more
-
please share your ideas in comments
Discussion
Comment from kucai on 5 July 2021 at 03:25
The assumption for that is all the gps traces are of equivalent quality, accuracy wise. Other things that need to be considered is the already existance of surveyed points, averaged points, imageries distortions etc..
Comment from Ltrlg on 5 July 2021 at 07:26
Having a database like Imagery Offset Database seems to be a better solution to misalignment than automated road/building/etc. detection alone. Automated detection can be used to fill the database, but having human curation in a common database means that editor have the same working offset. It is a bit useless without default support in editors though.
Comment from SimonPoole on 5 July 2021 at 13:52
@Ltrlg Zverik is both the author and operator of the imagery offset database, so I think he does know about it :-).
That said, I still think (this is not a new discussion) that the way to go is not to completely automate things but at least add support to iD (all other general purpose editors have support I believe), and at least suggest using offsets if they exist for the current imagery.
Comment from imagico on 5 July 2021 at 17:30
Yes, automation of mechanical work in mapping is severely underdeveloped in OSM at the moment.
But this concerns not so much what many people think of when they hear “automation in mapping in OSM”, i.e. the large scale industrial automation with low quality results a la Facebook where the mapper - if at all - only remains as a click-monkey to mechanically confirm and take responsibility for entering the data without actually contributing local knowledge. The unused potential is IMO more on the level of micro-automation of mechanical and time consuming steps in the work of mappers. Like for example line/polygon geometry generation of individual features identified by the mapper - the iconic example would be the one click mapping of buildings from imagery.
By the way - automation of imagery alignment could in addition or instead of using GPS traces (which can be tricky because assigning a GPS trace to a feature visible in an image often requires human judgement) be based on cross correlation between different imagery sources (in particular of lower resolution imagery of vertical view satellite imagery with known and low systematic position errors).