Urban Change Forecasting from Satellite Images

TL;DR

This paper introduces a deep learning approach for predicting urban development from satellite images, including when and where new buildings will appear, using a novel pretraining method that outperforms traditional techniques.

Contribution

The work presents a new pretraining procedure for a neural network to forecast urban changes from satellite images, including timing predictions, validated on the SpaceNet7 dataset.

Findings

Pretraining with the proposed method outperforms ImageNet pretraining.

The model can predict the timing of building changes to some extent.

Validated on a large-scale satellite dataset with promising results.

Abstract

Forecasting where and when new buildings will emerge is a rather unexplored topic, but one that is very useful in many disciplines such as urban planning, agriculture, resource management, and even autonomous flying. In the present work, we present a method that accomplishes this task with a deep neural network and a custom pretraining procedure. In Stage 1, a U-Net backbone is pretrained within a Siamese network architecture that aims to solve a (building) change detection task. In Stage 2, the backbone is repurposed to forecast the emergence of new buildings based solely on one image acquired before its construction. Furthermore, we also present a model that forecasts the time range within which the change will occur. We validate our approach using the SpaceNet7 dataset, which covers an area of 960 km^2 at 24 points in time across two years. In our experiments, we found that our proposed pretraining method consistently outperforms the traditional pretraining using the ImageNet dataset. We also show that it is to some degree possible to predict in advance when building changes will occur.