Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution Network

Antarctic true-color imagery synthesized using multispectral remote sensing data is effective in reflecting sea ice conditions, which is crucial for monitoring. Deep learning has been explored for sea ice extraction, but traditional convolutional neural network models are constrained by a limited pe...

Full description

Saved in:
Bibliographic Details
Main Authors: Wu Feng, Xiulin Geng, Xiaoyu He, Miao Hu, Jie Luo, Meihua Bi
Format: Article
Language:English
Published: MDPI AG 2025-02-01
Series:Journal of Marine Science and Engineering
Subjects:
Online Access:https://www.mdpi.com/2077-1312/13/3/439
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849342459539095552
author Wu Feng
Xiulin Geng
Xiaoyu He
Miao Hu
Jie Luo
Meihua Bi
author_facet Wu Feng
Xiulin Geng
Xiaoyu He
Miao Hu
Jie Luo
Meihua Bi
author_sort Wu Feng
collection DOAJ
description Antarctic true-color imagery synthesized using multispectral remote sensing data is effective in reflecting sea ice conditions, which is crucial for monitoring. Deep learning has been explored for sea ice extraction, but traditional convolutional neural network models are constrained by a limited perceptual field, making it difficult to obtain global contextual information from remote sensing images. A novel model named GEFU-Net, a modification of U-Net, is presented. The self-established graph reconstruction module is employed to convert features into graph data and construct the adjacency matrix using a global adaptive average similarity threshold. Graph convolutional networks are utilized to aggregate the features at each pixel, enabling the rapid capture of global context, enhancing the semantic richness of the features, and improving the accuracy of sea ice extraction through graph reconstruction. Experimental results using the sea ice dataset of the Ross Sea in the Antarctic, produced by Sentinel-2, demonstrate that our GEFU-Net achieves the best performance compared to other commonly used segmentation models. Specifically, it achieves an accuracy of 97.52%, an Intersection over Union of 95.66%, and an F1-Score of 97.78%. Additionally, fewer model parameters and good inference speed are demonstrated, indicating strong potential for practical ice mapping applications.
format Article
id doaj-art-7f5b3bd1ca604a4784cdea1e6e96705d
institution Kabale University
issn 2077-1312
language English
publishDate 2025-02-01
publisher MDPI AG
record_format Article
series Journal of Marine Science and Engineering
spelling doaj-art-7f5b3bd1ca604a4784cdea1e6e96705d2025-08-20T03:43:22ZengMDPI AGJournal of Marine Science and Engineering2077-13122025-02-0113343910.3390/jmse13030439Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution NetworkWu Feng0Xiulin Geng1Xiaoyu He2Miao Hu3Jie Luo4Meihua Bi5School of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, ChinaSchool of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, ChinaSchool of Computer Science and Technology, Zhejiang Sci-Tech University, Hangzhou 310018, ChinaSchool of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, ChinaSchool of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, ChinaSchool of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, ChinaAntarctic true-color imagery synthesized using multispectral remote sensing data is effective in reflecting sea ice conditions, which is crucial for monitoring. Deep learning has been explored for sea ice extraction, but traditional convolutional neural network models are constrained by a limited perceptual field, making it difficult to obtain global contextual information from remote sensing images. A novel model named GEFU-Net, a modification of U-Net, is presented. The self-established graph reconstruction module is employed to convert features into graph data and construct the adjacency matrix using a global adaptive average similarity threshold. Graph convolutional networks are utilized to aggregate the features at each pixel, enabling the rapid capture of global context, enhancing the semantic richness of the features, and improving the accuracy of sea ice extraction through graph reconstruction. Experimental results using the sea ice dataset of the Ross Sea in the Antarctic, produced by Sentinel-2, demonstrate that our GEFU-Net achieves the best performance compared to other commonly used segmentation models. Specifically, it achieves an accuracy of 97.52%, an Intersection over Union of 95.66%, and an F1-Score of 97.78%. Additionally, fewer model parameters and good inference speed are demonstrated, indicating strong potential for practical ice mapping applications.https://www.mdpi.com/2077-1312/13/3/439sea ice extractiontrue-color imagerysentinel-2 imagedeep learningfeature enhancementgraph convolution network
spellingShingle Wu Feng
Xiulin Geng
Xiaoyu He
Miao Hu
Jie Luo
Meihua Bi
Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution Network
Journal of Marine Science and Engineering
sea ice extraction
true-color imagery
sentinel-2 image
deep learning
feature enhancement
graph convolution network
title Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution Network
title_full Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution Network
title_fullStr Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution Network
title_full_unstemmed Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution Network
title_short Antarctic Sea Ice Extraction for Remote Sensing Images via Modified U-Net Based on Feature Enhancement Driven by Graph Convolution Network
title_sort antarctic sea ice extraction for remote sensing images via modified u net based on feature enhancement driven by graph convolution network
topic sea ice extraction
true-color imagery
sentinel-2 image
deep learning
feature enhancement
graph convolution network
url https://www.mdpi.com/2077-1312/13/3/439
work_keys_str_mv AT wufeng antarcticseaiceextractionforremotesensingimagesviamodifiedunetbasedonfeatureenhancementdrivenbygraphconvolutionnetwork
AT xiulingeng antarcticseaiceextractionforremotesensingimagesviamodifiedunetbasedonfeatureenhancementdrivenbygraphconvolutionnetwork
AT xiaoyuhe antarcticseaiceextractionforremotesensingimagesviamodifiedunetbasedonfeatureenhancementdrivenbygraphconvolutionnetwork
AT miaohu antarcticseaiceextractionforremotesensingimagesviamodifiedunetbasedonfeatureenhancementdrivenbygraphconvolutionnetwork
AT jieluo antarcticseaiceextractionforremotesensingimagesviamodifiedunetbasedonfeatureenhancementdrivenbygraphconvolutionnetwork
AT meihuabi antarcticseaiceextractionforremotesensingimagesviamodifiedunetbasedonfeatureenhancementdrivenbygraphconvolutionnetwork