In a constantly evolving urban landscape, it is particularly hard to measure the rate of growth of informal housing, which can constitute 30%‐60% of the city, surpassing, in some cases, the extent of formal neighborhoods. However, identifying irregular settlements, mapping and monitoring them using traditional approaches is often costly and labor intensive. In a bid to offer a solution to this problem, this study looks at using machine learning to identify informal housing through satellite imagery. Very high‐resolution multispectral satellite imagery has proven to be highly useful in settlement mapping; however, irregular settlement mapping is still an open challenge. This research was published by the Duke University as a companion to the paper titled ‘Studying the Real Slums of Bengaluru’.
KEY MESSAGES
- The actual number of slums is much higher than what is recorded in official government records.
- Change detection methods have shown a significant change in informal settlements over the last 15 years.
- Machine learning algorithms are better at detecting slums and informal housing at the lower end of the quality of housing spectrum. However, in cases of slums with better housing quality, the algorithm often confuses it with formal housing.