Domain-Adversarial Neural Networks for Deforestation Detection in Tropical Forests

Type Article
Date 2022
Language English
Author(s) Soto Vega Pedro Juan1, Costa Gilson A.2, Feitosa Raul Q.3, Ortega Mabel X.3, Bermudez Jose D.3, Turnes Javier N.4
Affiliation(s) 1 : French Research Institute for Exploitation of the Sea, Brest, France.
2 : State University of Rio de Janeiro, Rio de Janeiro, Brazil.
3 : Department of Electrical Engineering, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro, Brazil.
4 : Department of System Design Engineering, University of Waterloo, Waterloo, Canada
Source Ieee Geoscience And Remote Sensing Letters (1545-598X) (Institute of Electrical and Electronics Engineers (IEEE)), 2022 , Vol. 19 , N. 2504505 , P. 5p.
DOI 10.1109/LGRS.2022.3163575
WOS© Times Cited 2
Keyword(s) Feature extraction, Training, Neurons, Task analysis, Neural networks, Forestry, Noise measurement, Change detection, deep learning (DL), deforestation detection, domain adaptation (DA), remote sensing (RS)
Abstract

Many deep-learning-based, domain adaptation methods for remote sensing applications rely on adversarial training strategies to align features extracted from images of different domains in a shared latent space. However, the performance of such representation matching techniques is negatively impacted when class occurrences in the target domain, for which no labelled data is available during training, are highly imbalanced. In this work, we propose a deep-learning-based representation matching approach for domain adaptation in the context of change detection tasks. We further evaluate the approach in a deforestation mapping application, characterized by a high-class imbalance between the deforestation and no-deforestation classes. The domains represent different sites in the Amazon and Brazilian Cerrado biomes. To mitigate the class imbalance problem, we devised an unsupervised pseudo-labeling scheme based on Change Vector Analysis that prevents the feature alignment to be biased towards the over-represented class. The experimental results indicate that the proposed approach can improve the accuracy of cross-domain deforestation detection.

Full Text
File Pages Size Access
Author's final draft 6 2 MB Open access
5 2 MB Access on demand
Top of the page