Detecting urban changes with recurrent neural networks from multitemporal Sentinel-2 data - Archive ouverte HAL Access content directly
Conference Papers Year :

Detecting urban changes with recurrent neural networks from multitemporal Sentinel-2 data

(1, 2, 3, 4) , (3, 4) , (5, 3, 4) , (6) , (7)
1
2
3
4
5
6
7

Abstract

The advent of multitemporal high resolution data, like theCopernicus Sentinel-2, has enhanced significantly the poten-tial of monitoring the earth’s surface and environmental dy-namics. In this paper, we present a novel deep learning frame-work for urban change detection which combines state-of-the-art fully convolutional networks (similar to U-Net) forfeature representation and powerful recurrent networks (suchas LSTMs) for temporal modeling. We report our resultson the recently publicly available bi-temporal Onera Satel-lite Change Detection (OSCD) Sentinel-2 dataset, enhancingthe temporal information with additional images of the sameregion on different dates. Moreover, we evaluate the perfor-mance of the recurrent networks as well as the use of the ad-ditional dates on the unseen test-set using an ensemble cross-validation strategy. All the developed models during the val-idation phase have scored an overall accuracy of more than95%, while the use of LSTMs and further temporal informa-tion, boost the F1 rate of the change class by an additional 1,5%.
Fichier principal
Vignette du fichier
20190527122608_825039_4393.pdf (1.04 Mo) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02266094 , version 1 (13-08-2019)

Identifiers

  • HAL Id : hal-02266094 , version 1

Cite

Maria Papadomanolaki, Sagar Verma, Maria Vakalopoulou, Siddharth Gupta, Konstantinos Karantzalos. Detecting urban changes with recurrent neural networks from multitemporal Sentinel-2 data. IGARSS 2019 - IEEE International Geoscience and Remote Sensing Symposium, Jul 2019, Yokohama, Japan. ⟨hal-02266094⟩
158 View
589 Download

Share

Gmail Facebook Twitter LinkedIn More