Cloud detection on satellite imagery using U-Net architecture

Alexis Vandoit; Benoit Blanco; Léo Turon; Clément Killisly

doi:10.1117/12.2683412

19 October 2023 Cloud detection on satellite imagery using U-Net architecture

Alexis Vandoit, Benoit Blanco, Léo Turon, Clément Killisly

Proceedings Volume 12730, Remote Sensing of Clouds and the Atmosphere XXVIII; 1273005 (2023) https://doi.org/10.1117/12.2683412
Event: SPIE Remote Sensing, 2023, Amsterdam, Netherlands

Abstract

Cloud detection holds significant importance when analyzing satellite imagery, notably in the visible domain. A wide variety of detection tools already exists for this type of application, using radiometric information. With the aim of enhancing cloud detection, recent advances in deep learning have made it possible to create tools based on pattern recognition while leveraging radiometric data. This paper aims to present a method focused on machine learning, giving details of its construction process, from the creation of the datasets to overall performance. A sample use case is also presented to demonstrate the promising outcomes obtained. Using over 600,000km² of ground-labeled satellite imagery and a U-Net based architecture for our machine learning algorithm, we achieved encouraging performances over various land types. Our results showed significant results when evaluated on the three most frequently used metrics for Image Segmentation. Our product gives interesting results in certain areas that present challenging ground types, such as snowy tops, when using Infrared imagery.

Conference Presentation

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Alexis Vandoit, Benoit Blanco, Léo Turon, and Clément Killisly "Cloud detection on satellite imagery using U-Net architecture", Proc. SPIE 12730, Remote Sensing of Clouds and the Atmosphere XXVIII, 1273005 (19 October 2023); https://doi.org/10.1117/12.2683412

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available