londonmapbot
londonmapbot copied to clipboard
matt-dray
Check if point is within Greater London boundary
Bit of extra effort, but perhaps you could store the boundary of Greater London (you can pull them from the ONS Geoportal using sf::st_read(https://opendata.arcgis.com/datasets/324a9f3ad2cc4a049e913dddc508aeb2_0.geojson?where=rgn19nm%20%3D%20'London')) and then check whether the random points calculated are located within this polygon.
True.
This is the type of information that could be posted with the tweet itself, along with things like nearest tube (as per Ken Butler). I like the simplicity of the tweet content for now.
I'll definitely be doing further analysis, though. For example, a map app with pop-up info (like in/out of Greater London, what borough it's in, etc). Might try training a classifier on 'rural'/'urban' too.
Agree with simplicity, was mainly thinking about it only giving "valid" London coordinates.
On rural/urban there's an ONS rural/urban classification.
Haha, I know all too well: I was on the team that devised and developed that classification! I was thinking of hand-tagging with a simple binary of 'built-up' versus not and then predicting new tweets as they come in.
Would this risk tagging photos that are bang in the middle of large parks (Hyde Park, Richmond Park, Hampstead Heath) as "rural"? Maybe a different description "green/grey/brown":
- Green: park/farmland
- Grey: built environment
- Brown: (ex-)industrial/construction
Also how it would treat large bodies of water (eg reservoirs and the lower reaches of the Thames)?
Water isn't built-up, so it would be 'not built-up' 😄
I was thinking of doing it for a 'train a classifier for beginners' sort of blog post, so simplicity would be key. Could demonstrate a few approaches re binary/multi-class. There's definitely a question about images that have multiple land-use types. Could quantify colour in the image to guess the predominant case.
Related: a post from Ken Butler on sampling points from within Toronto limits: http://ritsokiguess.site/docs/2020/10/10/sampling-from-locations-in-a-city/