Multi-nomenclature, multi-resolution joint translation: an application to land-cover mapping

Abstract

Land-use/land-cover (LULC) maps describe the Earth’s surface with discrete classes at a specific spatial resolution. The chosen classes and resolution highly depend on peculiar uses, making it mandatory to develop methods to adapt these characteristics for a large range of applications. Recently, a convolutional neural network (CNN)-based method was introduced to take into account both spatial and geographical context to translate a LULC map into another one. However, this model only works for two maps: one source and one target. Inspired by natural language translation using multiple-language models, this article explores how to translate one LULC map into several targets with distinct nomenclatures and spatial resolutions. We first propose a new data set based on six open access LULC maps to train our CNN-based encoder-decoder framework. We then apply such a framework to convert each of these six maps into each of the others using our Multi-Landcover Translation network (MLCT-Net). Extensive experiments are conducted at a country scale (namely France). The results reveal that our MLCT-Net outperforms its semantic counterparts and gives on par results with mono-LULC models when evaluated on areas similar to those used for training. Furthermore, it outperforms the mono-LULC models when applied to totally new landscapes.

Keywords:

• Land-cover
• land-use
• translation
• deep learning
• harmonization

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data and codes availability statement

Data are made available at https://doi.org/10.5281/zenodo.5843595 and code at https://doi.org/10.5281/zenodo.7019838.

Additional information

Funding

This work is funded by the MAESTRIA project [grant ANR-18-CE23-0023] and supported by the AI4GEO project http://www.ai4geo.eu/ and Agence Nationale de la Recherche.

Notes on contributors

Luc Baudoux

Luc Baudoux is a PhD candidate at the LASTIG and CESBIO laboratory at the Gustave Eiffel University in France. His research investigates how to increase land-cover quality and reusability.

Jordi Inglada

Jordi Inglada Jordi Inglada received the master’s degree in telecommunications engineering from the Universitat Politècnica de Catalunya, Barcelona, Spain, and the École Nationale Supérieure des Télécommunications de Bretagne, Brest, France, in 1997, and the Ph.D. degree in signalprocessing and telecommunications from the Université de Rennes 1, Rennes, France, in 2000. He is currently with the Centre National d’Etudes Spatiales (French Space Agency), Toulouse, France, where he is involved in the field of remote sensing image processing at the Centre d’Etudes Spatiales de la Biosphère (CESBIO) Laboratory. He is involved in the development of image processing algorithms for the operational exploitation of Earth observation images, mainly in the field of multitemporal image analysis for land use and cover change.

Clément Mallet

Clément Mallet is currently leading the LASTIG laboratory (Univ. Gustave Eiffel and French Mapping Agency) and is Editor-in-Chief of the ISPRS Journal of Photogrammetry and Remote Sensing. His main interests are land-cover mapping with remote sensing imagery and point cloud processing.