Atmospheric Correction of Sentinel-2 Images Using Deep Learning

Stavan Shah, Kahaan Patel, Ankur Garg, Shivangi Surati

2024

Abstract

Remote sensing relies heavily on pre-processing steps, one of which is the Atmospheric Correction (AC). It corrects the effects of atmosphere on satellite images. This makes it a vital step in ensuring accurate estimation of land Surface Reflectance (SR) that can be used in various downstream applications. But such conventional AC methods are computationally expensive because they use physics-based radiative transfer codes, need metadata from each image as well as many different atmospheric parameters which might not all be easy to estimate accurately. A novel Deep Learning (DL) model designed for AC without having to explicitly estimate atmospheric parameters is proposed in this research. The deep learning model was trained using a wide-ranging dataset collected by Google Earth Engine that included four bands of Sentinel 2 images covering all states in India. The proposed approach directly predicts SR values from Sentinel-2 satellite imagery using this data driven method. It generated promising results by accurately estimating SR values with ground measurements and sentinel input data experiments confirming this point too. This approach not only simplifies the AC process but also achieves comparable or even superior performance compared to traditional physics-based methods. The evaluation results show that Pix2Pix model has good performance, with average SSIM, PSNR, RMSE and MAE of 0.96, 42.14, 0.0097 and 0.0071 respectively. The experimental findings underscore the potential of deep learning as a robust and efficient alternative for atmospheric correction in remote sensing applications.

Paper Citation

in Harvard Style

Shah S., Patel K., Garg A. and Surati S. (2024). Atmospheric Correction of Sentinel-2 Images Using Deep Learning. In Proceedings of the 1st International Conference on Cognitive & Cloud Computing - Volume 1: IC3Com; ISBN 978-989-758-739-9, SciTePress, pages 175-185. DOI: 10.5220/0013303900004646

in Bibtex Style

@conference{ic3com24,
author={Stavan Shah and Kahaan Patel and Ankur Garg and Shivangi Surati},
title={Atmospheric Correction of Sentinel-2 Images Using Deep Learning},
booktitle={Proceedings of the 1st International Conference on Cognitive & Cloud Computing - Volume 1: IC3Com},
year={2024},
pages={175-185},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013303900004646},
isbn={978-989-758-739-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Cognitive & Cloud Computing - Volume 1: IC3Com
TI - Atmospheric Correction of Sentinel-2 Images Using Deep Learning
SN - 978-989-758-739-9
AU - Shah S.
AU - Patel K.
AU - Garg A.
AU - Surati S.
PY - 2024
SP - 175
EP - 185
DO - 10.5220/0013303900004646
PB - SciTePress