Self-supervised pre-training enhances change detection in Sentinel-2 imagery

TL;DR

This paper demonstrates that self-supervised pre-training using Sentinel-2 multitemporal imagery improves change detection performance, leveraging a new large dataset and temporal consistency for better representations.

Contribution

It introduces a novel self-supervised learning approach for satellite image change detection and provides a new publicly available dataset, S2MTCP, for this task.

Findings

Self-supervised pre-training enhances change detection accuracy.

The S2MTCP dataset enables effective model training.

Results outperform baseline methods on the OSCD dataset.

Abstract

While annotated images for change detection using satellite imagery are scarce and costly to obtain, there is a wealth of unlabeled images being generated every day. In order to leverage these data to learn an image representation more adequate for change detection, we explore methods that exploit the temporal consistency of Sentinel-2 times series to obtain a usable self-supervised learning signal. For this, we build and make publicly available (https://zenodo.org/record/4280482) the Sentinel-2 Multitemporal Cities Pairs (S2MTCP) dataset, containing multitemporal image pairs from 1520 urban areas worldwide. We test the results of multiple self-supervised learning methods for pre-training models for change detection and apply it on a public change detection dataset made of Sentinel-2 image pairs (OSCD).