Detecting urban typology from multispectral satellite imagery using neural networks

Martin Fleischmann
Dani Arribas-Bel
John Murray
Alex Singleton

The data

A quick overview of data used in the recent publications in Urban Morphology.

Fu et al. (2022)

Kantarek et al. (2022)

Guo and Ding (2021)

Li and Zhang (2021)

The issue

  • availability
  • reliability
  • processing demands

The issue

Urban morphology is bounded by the data availability and the ability to extract morphological information out of it.

The solution (?)

 🛰

Sentinel 2

Morphology and imagery

  • supervised methods
  • unsupervised methods

Predicting Spatial Signatures

supervised learning

Spatial Signatures

A characterisation of space based on form and function designed to understand urban environments

What do we want to do?

train a neural network
understand the role of geography

Chip size effect

80x80m: 13760 chips, which is 74 % of maximum

160x160m: 2718 chips within, which is 57 % of maximum

320x320m: 423 chips within, which is 35 % of maximum

640x640m: 38 chips within, which is 13 % of maximum

320x320m, chips capturing the proportion (100% of maximum)

Predicted class (320x320m)

Wild countryside (320x320m)

Urbanity (320x320m)

Still work in progress

Estimating generalized measures of local neighbourhood context

unsupervised learning

The takeaway

  • We are limited by the data we are used to
  • Can satellite imagery and AI resolve it?
  • Probably not. Not all of it and not yet.

Detecting urban typology from multispectral satellite imagery using neural networks

Martin Fleischmann
Dani Arribas-Bel
John Murray
Alex Singleton