Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes
Polar mesospheric summer echoes (PMSE) are radar echoes that are observed in the mesosphere during the arctic summer months in the polar regions. By studying PMSE, researchers can gain insights into physical and chemical processes that occur in the upper atmosphere—specifically, in the 80 to 90 km a...
Published in: | Remote Sensing |
---|---|
Main Authors: | , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
MDPI AG
2023
|
Subjects: | |
Online Access: | https://doi.org/10.3390/rs15174291 https://doaj.org/article/9297871a6dfd445fb9e0a5957fca2f50 |
id |
ftdoajarticles:oai:doaj.org/article:9297871a6dfd445fb9e0a5957fca2f50 |
---|---|
record_format |
openpolar |
spelling |
ftdoajarticles:oai:doaj.org/article:9297871a6dfd445fb9e0a5957fca2f50 2023-10-09T21:48:59+02:00 Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes Erik Seip Domben Puneet Sharma Ingrid Mann 2023-08-01T00:00:00Z https://doi.org/10.3390/rs15174291 https://doaj.org/article/9297871a6dfd445fb9e0a5957fca2f50 EN eng MDPI AG https://www.mdpi.com/2072-4292/15/17/4291 https://doaj.org/toc/2072-4292 doi:10.3390/rs15174291 2072-4292 https://doaj.org/article/9297871a6dfd445fb9e0a5957fca2f50 Remote Sensing, Vol 15, Iss 4291, p 4291 (2023) polar mesospheric summer echoes deep learning segmentation Science Q article 2023 ftdoajarticles https://doi.org/10.3390/rs15174291 2023-09-10T00:34:43Z Polar mesospheric summer echoes (PMSE) are radar echoes that are observed in the mesosphere during the arctic summer months in the polar regions. By studying PMSE, researchers can gain insights into physical and chemical processes that occur in the upper atmosphere—specifically, in the 80 to 90 km altitude range. In this paper, we employ fully convolutional networks such as UNET and UNET++ for the purpose of segmenting PMSE from the EISCAT VHF dataset. First, experiments are performed to find suitable weights and hyperparameters for UNET and UNET++. Second, different loss functions are tested to find one suitable for our task. Third, as the number of PMSE samples used is relatively small, this can lead to poor generalization. To address this, image-level and object-level augmentation methods are employed. Fourth, we briefly explain our findings by employing layerwise relevance propagation. Article in Journal/Newspaper Arctic EISCAT Directory of Open Access Journals: DOAJ Articles Arctic Remote Sensing 15 17 4291 |
institution |
Open Polar |
collection |
Directory of Open Access Journals: DOAJ Articles |
op_collection_id |
ftdoajarticles |
language |
English |
topic |
polar mesospheric summer echoes deep learning segmentation Science Q |
spellingShingle |
polar mesospheric summer echoes deep learning segmentation Science Q Erik Seip Domben Puneet Sharma Ingrid Mann Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes |
topic_facet |
polar mesospheric summer echoes deep learning segmentation Science Q |
description |
Polar mesospheric summer echoes (PMSE) are radar echoes that are observed in the mesosphere during the arctic summer months in the polar regions. By studying PMSE, researchers can gain insights into physical and chemical processes that occur in the upper atmosphere—specifically, in the 80 to 90 km altitude range. In this paper, we employ fully convolutional networks such as UNET and UNET++ for the purpose of segmenting PMSE from the EISCAT VHF dataset. First, experiments are performed to find suitable weights and hyperparameters for UNET and UNET++. Second, different loss functions are tested to find one suitable for our task. Third, as the number of PMSE samples used is relatively small, this can lead to poor generalization. To address this, image-level and object-level augmentation methods are employed. Fourth, we briefly explain our findings by employing layerwise relevance propagation. |
format |
Article in Journal/Newspaper |
author |
Erik Seip Domben Puneet Sharma Ingrid Mann |
author_facet |
Erik Seip Domben Puneet Sharma Ingrid Mann |
author_sort |
Erik Seip Domben |
title |
Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes |
title_short |
Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes |
title_full |
Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes |
title_fullStr |
Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes |
title_full_unstemmed |
Using Deep Learning Methods for Segmenting Polar Mesospheric Summer Echoes |
title_sort |
using deep learning methods for segmenting polar mesospheric summer echoes |
publisher |
MDPI AG |
publishDate |
2023 |
url |
https://doi.org/10.3390/rs15174291 https://doaj.org/article/9297871a6dfd445fb9e0a5957fca2f50 |
geographic |
Arctic |
geographic_facet |
Arctic |
genre |
Arctic EISCAT |
genre_facet |
Arctic EISCAT |
op_source |
Remote Sensing, Vol 15, Iss 4291, p 4291 (2023) |
op_relation |
https://www.mdpi.com/2072-4292/15/17/4291 https://doaj.org/toc/2072-4292 doi:10.3390/rs15174291 2072-4292 https://doaj.org/article/9297871a6dfd445fb9e0a5957fca2f50 |
op_doi |
https://doi.org/10.3390/rs15174291 |
container_title |
Remote Sensing |
container_volume |
15 |
container_issue |
17 |
container_start_page |
4291 |
_version_ |
1779312028794486784 |