End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning

Giorgio Morales; Alejandro Ramirez; Joel Telles

doi:10.1109/INTERCON.2019.8853549

End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning

Giorgio Morales^*
, Alejandro Ramirez
, Joel Telles

^*Corresponding author for this work

National University of Engineering

Research output: Chapter in Book/Published conference output › Conference publication

13 Citations (Scopus)

Abstract

Segmenting clouds in high-resolution satellite images is an arduous and challenging task due to the many types of geographies and clouds a satellite can capture. Therefore, it needs to be automated and optimized, specially for those who regularly process great amounts of satellite images, such as governmental institutions. In that sense, the contribution of this work is twofold: We present the CloudPeru2 dataset, consisting of 22,400 images of 512 × 512 pixels and their respective hand-drawn cloud masks, as well as the proposal of an end-to-end segmentation method for clouds using a Convolutional Neural Network (CNN) based on the Deeplab v3+ architecture. The results over the test set achieved an accuracy of 96.62%, precision of 96.46%, specificity of 98.53%, and sensitivity of 96.72% which is superior to the compared methods.

Original language	English
Title of host publication	Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019
Publisher	IEEE
ISBN (Electronic)	9781728136462
DOIs	https://doi.org/10.1109/INTERCON.2019.8853549
Publication status	Published - 3 Oct 2019
Event	26th IEEE International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019 - Lima, Peru Duration: 12 Aug 2019 → 14 Aug 2019

Publication series

Name	Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019

Conference

Conference	26th IEEE International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019
Country/Territory	Peru
City	Lima
Period	12/08/19 → 14/08/19

Bibliographical note

Publisher Copyright:
© 2019 IEEE.

Keywords

Cloud segmentation
End-to-end learning
Satellite image

Access to Document

10.1109/INTERCON.2019.8853549

https://arxiv.org/abs/1904.12743Licence: CC BY-NC-SA 4.0

Cite this

Morales, G., Ramirez, A., & Telles, J. (2019). End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning. In Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019 Article 8853549 (Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019). IEEE. https://doi.org/10.1109/INTERCON.2019.8853549

Morales, Giorgio ; Ramirez, Alejandro ; Telles, Joel. / End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning. Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019. IEEE, 2019. (Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019).

@inproceedings{6f0a391f836641b0969db9a8e65d6bff,

title = "End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning",

abstract = "Segmenting clouds in high-resolution satellite images is an arduous and challenging task due to the many types of geographies and clouds a satellite can capture. Therefore, it needs to be automated and optimized, specially for those who regularly process great amounts of satellite images, such as governmental institutions. In that sense, the contribution of this work is twofold: We present the CloudPeru2 dataset, consisting of 22,400 images of 512 × 512 pixels and their respective hand-drawn cloud masks, as well as the proposal of an end-to-end segmentation method for clouds using a Convolutional Neural Network (CNN) based on the Deeplab v3+ architecture. The results over the test set achieved an accuracy of 96.62\%, precision of 96.46\%, specificity of 98.53\%, and sensitivity of 96.72\% which is superior to the compared methods.",

keywords = "Cloud segmentation, End-to-end learning, Satellite image",

author = "Giorgio Morales and Alejandro Ramirez and Joel Telles",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 26th IEEE International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019 ; Conference date: 12-08-2019 Through 14-08-2019",

year = "2019",

month = oct,

day = "3",

doi = "10.1109/INTERCON.2019.8853549",

language = "English",

series = "Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019",

publisher = "IEEE",

booktitle = "Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019",

address = "United States",

}

Morales, G, Ramirez, A & Telles, J 2019, End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning. in Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019., 8853549, Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019, IEEE, 26th IEEE International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019, Lima, Peru, 12/08/19. https://doi.org/10.1109/INTERCON.2019.8853549

End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning. / Morales, Giorgio; Ramirez, Alejandro; Telles, Joel.
Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019. IEEE, 2019. 8853549 (Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019).

Research output: Chapter in Book/Published conference output › Conference publication

TY - GEN

T1 - End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning

AU - Morales, Giorgio

AU - Ramirez, Alejandro

AU - Telles, Joel

PY - 2019/10/3

Y1 - 2019/10/3

N2 - Segmenting clouds in high-resolution satellite images is an arduous and challenging task due to the many types of geographies and clouds a satellite can capture. Therefore, it needs to be automated and optimized, specially for those who regularly process great amounts of satellite images, such as governmental institutions. In that sense, the contribution of this work is twofold: We present the CloudPeru2 dataset, consisting of 22,400 images of 512 × 512 pixels and their respective hand-drawn cloud masks, as well as the proposal of an end-to-end segmentation method for clouds using a Convolutional Neural Network (CNN) based on the Deeplab v3+ architecture. The results over the test set achieved an accuracy of 96.62%, precision of 96.46%, specificity of 98.53%, and sensitivity of 96.72% which is superior to the compared methods.

AB - Segmenting clouds in high-resolution satellite images is an arduous and challenging task due to the many types of geographies and clouds a satellite can capture. Therefore, it needs to be automated and optimized, specially for those who regularly process great amounts of satellite images, such as governmental institutions. In that sense, the contribution of this work is twofold: We present the CloudPeru2 dataset, consisting of 22,400 images of 512 × 512 pixels and their respective hand-drawn cloud masks, as well as the proposal of an end-to-end segmentation method for clouds using a Convolutional Neural Network (CNN) based on the Deeplab v3+ architecture. The results over the test set achieved an accuracy of 96.62%, precision of 96.46%, specificity of 98.53%, and sensitivity of 96.72% which is superior to the compared methods.

KW - Cloud segmentation

KW - End-to-end learning

KW - Satellite image

UR - https://www.scopus.com/pages/publications/85073569622

UR - https://ieeexplore.ieee.org/document/8853549

U2 - 10.1109/INTERCON.2019.8853549

DO - 10.1109/INTERCON.2019.8853549

M3 - Conference publication

AN - SCOPUS:85073569622

T3 - Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019

BT - Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019

PB - IEEE

T2 - 26th IEEE International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019

Y2 - 12 August 2019 through 14 August 2019

ER -

Morales G, Ramirez A, Telles J. End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning. In Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019. IEEE. 2019. 8853549. (Proceedings of the 2019 IEEE 26th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2019). doi: 10.1109/INTERCON.2019.8853549

End-to-end cloud segmentation in high-resolution multispectral satellite imagery using deep learning

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this