Work

RGB2LST: Enhancing Deep Learning-Based Land Surface Temperature Estimation with Multi-Modality and Artifacts Removal

Year: 2024

Type: conference-paper

Abstract: Accurate Land Surface Temperature (LST) estimation is crucial for understanding environmental dynamics and addressing diverse scientific and societal challenges. This study explores a novel approach for LST estimation using RGB data and integrating it with additional data modalities. By leveraging conditioned Generative Adversarial Networks (cGANs) for LST generation on adjacent tiles, certain artifacts are observed. To address this issue, we introduce a comprehensive processing pipeline and tiling strategy, evaluating fusion methods for artifact removal and improving LST generation accuracy. Our findings demonstrate the potential of integrating multi-modal data to enhance LST estimation, leading to promising advancements, particularly in LST data inference.

Source: European Signal Processing Conference (EUSIPCO)

DOI: 10.23919/eusipco63174.2024.10715154

Authors: Issam Khedher, Jean-Marie Favreau, Serge Miguet, Gilles Gesquière

Institutions: École Centrale de Lyon, Centre National de la Recherche Scientifique, Institut National des Sciences Appliquées de Lyon, Université Clermont Auvergne, Laboratoire d'Informatique, de Modélisation et d'Optimisation des Systèmes

Language: English

Topic: Urban Heat Island Mitigation

Subfield: Environmental Engineering

Field: Environmental Science

Domain: Physical Sciences

SDG: Climate action

Open Access status: green

Funders: Agence Nationale de la Recherche, Université de Lyon