Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks
We explore new and existing convolutional neural network (CNN) architectures for sea ice classification using Sentinel-1 (S1) synthetic aperture radar (SAR) data by investigating two key challenges: binary sea ice versus open-water classification, and a multi-class sea ice type classification. The a...
Published in: | Remote Sensing |
---|---|
Main Authors: | , , , , , |
Format: | Text |
Language: | English |
Published: |
Multidisciplinary Digital Publishing Institute
2021
|
Subjects: | |
Online Access: | https://doi.org/10.3390/rs13091734 |
id |
ftmdpi:oai:mdpi.com:/2072-4292/13/9/1734/ |
---|---|
record_format |
openpolar |
spelling |
ftmdpi:oai:mdpi.com:/2072-4292/13/9/1734/ 2023-08-20T04:09:41+02:00 Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks Salman Khaleghian Habib Ullah Thomas Kræmer Nick Hughes Torbjørn Eltoft Andrea Marinoni agris 2021-04-29 application/pdf https://doi.org/10.3390/rs13091734 EN eng Multidisciplinary Digital Publishing Institute Remote Sensing Image Processing https://dx.doi.org/10.3390/rs13091734 https://creativecommons.org/licenses/by/4.0/ Remote Sensing; Volume 13; Issue 9; Pages: 1734 convolutional neural network ice edge detection polar region Sentinel-1 sea ice classification synthetic aperture radar Text 2021 ftmdpi https://doi.org/10.3390/rs13091734 2023-08-01T01:37:22Z We explore new and existing convolutional neural network (CNN) architectures for sea ice classification using Sentinel-1 (S1) synthetic aperture radar (SAR) data by investigating two key challenges: binary sea ice versus open-water classification, and a multi-class sea ice type classification. The analysis of sea ice in SAR images is challenging because of the thermal noise effects and ambiguities in the radar backscatter for certain conditions that include the reflection of complex information from sea ice surfaces. We use manually annotated SAR images containing various sea ice types to construct a dataset for our Deep Learning (DL) analysis. To avoid contamination between classes we use a combination of near-simultaneous SAR images from S1 and fine resolution cloud-free optical data from Sentinel-2 (S2). For the classification, we use data augmentation to adjust for the imbalance of sea ice type classes in the training data. The SAR images are divided into small patches which are processed one at a time. We demonstrate that the combination of data augmentation and training of a proposed modified Visual Geometric Group 16-layer (VGG-16) network, trained from scratch, significantly improves the classification performance, compared to the original VGG-16 model and an ad hoc CNN model. The experimental results show both qualitatively and quantitatively that our models produce accurate classification results. Text Sea ice MDPI Open Access Publishing Remote Sensing 13 9 1734 |
institution |
Open Polar |
collection |
MDPI Open Access Publishing |
op_collection_id |
ftmdpi |
language |
English |
topic |
convolutional neural network ice edge detection polar region Sentinel-1 sea ice classification synthetic aperture radar |
spellingShingle |
convolutional neural network ice edge detection polar region Sentinel-1 sea ice classification synthetic aperture radar Salman Khaleghian Habib Ullah Thomas Kræmer Nick Hughes Torbjørn Eltoft Andrea Marinoni Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks |
topic_facet |
convolutional neural network ice edge detection polar region Sentinel-1 sea ice classification synthetic aperture radar |
description |
We explore new and existing convolutional neural network (CNN) architectures for sea ice classification using Sentinel-1 (S1) synthetic aperture radar (SAR) data by investigating two key challenges: binary sea ice versus open-water classification, and a multi-class sea ice type classification. The analysis of sea ice in SAR images is challenging because of the thermal noise effects and ambiguities in the radar backscatter for certain conditions that include the reflection of complex information from sea ice surfaces. We use manually annotated SAR images containing various sea ice types to construct a dataset for our Deep Learning (DL) analysis. To avoid contamination between classes we use a combination of near-simultaneous SAR images from S1 and fine resolution cloud-free optical data from Sentinel-2 (S2). For the classification, we use data augmentation to adjust for the imbalance of sea ice type classes in the training data. The SAR images are divided into small patches which are processed one at a time. We demonstrate that the combination of data augmentation and training of a proposed modified Visual Geometric Group 16-layer (VGG-16) network, trained from scratch, significantly improves the classification performance, compared to the original VGG-16 model and an ad hoc CNN model. The experimental results show both qualitatively and quantitatively that our models produce accurate classification results. |
format |
Text |
author |
Salman Khaleghian Habib Ullah Thomas Kræmer Nick Hughes Torbjørn Eltoft Andrea Marinoni |
author_facet |
Salman Khaleghian Habib Ullah Thomas Kræmer Nick Hughes Torbjørn Eltoft Andrea Marinoni |
author_sort |
Salman Khaleghian |
title |
Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks |
title_short |
Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks |
title_full |
Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks |
title_fullStr |
Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks |
title_full_unstemmed |
Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks |
title_sort |
sea ice classification of sar imagery based on convolution neural networks |
publisher |
Multidisciplinary Digital Publishing Institute |
publishDate |
2021 |
url |
https://doi.org/10.3390/rs13091734 |
op_coverage |
agris |
genre |
Sea ice |
genre_facet |
Sea ice |
op_source |
Remote Sensing; Volume 13; Issue 9; Pages: 1734 |
op_relation |
Remote Sensing Image Processing https://dx.doi.org/10.3390/rs13091734 |
op_rights |
https://creativecommons.org/licenses/by/4.0/ |
op_doi |
https://doi.org/10.3390/rs13091734 |
container_title |
Remote Sensing |
container_volume |
13 |
container_issue |
9 |
container_start_page |
1734 |
_version_ |
1774723289313181696 |