Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training
Nowadays, satellite image time series (SITS) are commonly employed to derive land-cover maps (LCM) to support decision makers in a variety of land management applications. In the most general workflow, the production of LCM strongly relies on available GT data to train supervised machine learning mo...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2023-01-01
|
Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/10089508/ |
_version_ | 1797330423781523456 |
---|---|
author | Emmanuel Capliez Dino Ienco Raffaele Gaetano Nicolas Baghdadi Adrien Hadj Salah |
author_facet | Emmanuel Capliez Dino Ienco Raffaele Gaetano Nicolas Baghdadi Adrien Hadj Salah |
author_sort | Emmanuel Capliez |
collection | DOAJ |
description | Nowadays, satellite image time series (SITS) are commonly employed to derive land-cover maps (LCM) to support decision makers in a variety of land management applications. In the most general workflow, the production of LCM strongly relies on available GT data to train supervised machine learning models. Unfortunately, these data are not always available due to time-consuming and costly field campaigns. In this scenario, the possibility to transfer a model learnt on a particular year (<italic>source domain</italic>) to a successive period of time (<italic>target domain</italic>), over the same study area, can save time and money. Such a kind of model transfer is challenging due to different acquisition conditions affecting each time period, thus resulting in possible distribution shifts between <italic>source</italic> and <italic>target</italic> domains. In the general field of machine learning, unsupervised domain adaptation (UDA) approaches are well suited to cope with the learning of models under distribution shifts between <italic>source</italic> and <italic>target</italic> domains. While widely explored in the general computer vision field, they are still underinvestigated for SITS-based land-cover mapping, especially for the temporal transfer scenario. With the aim to cope with this scenario in the context of SITS-based land-cover mapping, here we propose spatially aligned domain-adversarial neural network, a framework that combines both adversarial learning and self-training to transfer a classification model from a time period (year) to a successive one on a specific study area. Experimental assessment on a study area located in Burkina Faso characterized by challenging operational constraints demonstrates the significance of our proposal. The obtained results have shown that our proposal outperforms all the UDA competing methods by 7 to 12 points of <italic>F</italic>1-score across three different transfer tasks. |
first_indexed | 2024-03-08T07:19:38Z |
format | Article |
id | doaj.art-b2f732ecb8404c5bbfd48bc2b7c4da1c |
institution | Directory Open Access Journal |
issn | 2151-1535 |
language | English |
last_indexed | 2024-03-08T07:19:38Z |
publishDate | 2023-01-01 |
publisher | IEEE |
record_format | Article |
series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
spelling | doaj.art-b2f732ecb8404c5bbfd48bc2b7c4da1c2024-02-03T00:00:47ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352023-01-01163645367510.1109/JSTARS.2023.326375510089508Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-TrainingEmmanuel Capliez0https://orcid.org/0009-0001-8651-2183Dino Ienco1https://orcid.org/0000-0002-8736-3132Raffaele Gaetano2Nicolas Baghdadi3Adrien Hadj Salah4INRAE, UMR TETIS, University of Montpellier, Montpellier, FranceINRAE, UMR TETIS, University of Montpellier, Montpellier, FranceCIRAD, UMR TETIS, University of Montpellier, Montpellier, FranceINRAE, UMR TETIS, University of Montpellier, Montpellier, FranceAirbus Defence and Space, Toulouse, FranceNowadays, satellite image time series (SITS) are commonly employed to derive land-cover maps (LCM) to support decision makers in a variety of land management applications. In the most general workflow, the production of LCM strongly relies on available GT data to train supervised machine learning models. Unfortunately, these data are not always available due to time-consuming and costly field campaigns. In this scenario, the possibility to transfer a model learnt on a particular year (<italic>source domain</italic>) to a successive period of time (<italic>target domain</italic>), over the same study area, can save time and money. Such a kind of model transfer is challenging due to different acquisition conditions affecting each time period, thus resulting in possible distribution shifts between <italic>source</italic> and <italic>target</italic> domains. In the general field of machine learning, unsupervised domain adaptation (UDA) approaches are well suited to cope with the learning of models under distribution shifts between <italic>source</italic> and <italic>target</italic> domains. While widely explored in the general computer vision field, they are still underinvestigated for SITS-based land-cover mapping, especially for the temporal transfer scenario. With the aim to cope with this scenario in the context of SITS-based land-cover mapping, here we propose spatially aligned domain-adversarial neural network, a framework that combines both adversarial learning and self-training to transfer a classification model from a time period (year) to a successive one on a specific study area. Experimental assessment on a study area located in Burkina Faso characterized by challenging operational constraints demonstrates the significance of our proposal. The obtained results have shown that our proposal outperforms all the UDA competing methods by 7 to 12 points of <italic>F</italic>1-score across three different transfer tasks.https://ieeexplore.ieee.org/document/10089508/Deep learningland-cover mappingsatellite image time series (SITS)temporal-domain adaptation |
spellingShingle | Emmanuel Capliez Dino Ienco Raffaele Gaetano Nicolas Baghdadi Adrien Hadj Salah Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Deep learning land-cover mapping satellite image time series (SITS) temporal-domain adaptation |
title | Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training |
title_full | Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training |
title_fullStr | Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training |
title_full_unstemmed | Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training |
title_short | Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training |
title_sort | temporal domain adaptation for satellite image time series land cover mapping with adversarial learning and spatially aware self training |
topic | Deep learning land-cover mapping satellite image time series (SITS) temporal-domain adaptation |
url | https://ieeexplore.ieee.org/document/10089508/ |
work_keys_str_mv | AT emmanuelcapliez temporaldomainadaptationforsatelliteimagetimeserieslandcovermappingwithadversariallearningandspatiallyawareselftraining AT dinoienco temporaldomainadaptationforsatelliteimagetimeserieslandcovermappingwithadversariallearningandspatiallyawareselftraining AT raffaelegaetano temporaldomainadaptationforsatelliteimagetimeserieslandcovermappingwithadversariallearningandspatiallyawareselftraining AT nicolasbaghdadi temporaldomainadaptationforsatelliteimagetimeserieslandcovermappingwithadversariallearningandspatiallyawareselftraining AT adrienhadjsalah temporaldomainadaptationforsatelliteimagetimeserieslandcovermappingwithadversariallearningandspatiallyawareselftraining |