Fine extraction of arctic sea ice based on CA-DeepLabV3+ model

Shichang Sun; Zhiyong Wang; Kang Tian

doi:10.1117/12.3007422

30 August 2023 Fine extraction of arctic sea ice based on CA-DeepLabV3+ model

Shichang Sun, Zhiyong Wang, Kang Tian

Proceedings Volume 12797, Second International Conference on Geographic Information and Remote Sensing Technology (GIRST 2023); 127971P (2023) https://doi.org/10.1117/12.3007422
Event: 2nd International Conference on Geographic Information and Remote Sensing Technology (GIRST 2023), 2023, Qingdao, China

Abstract

Sea ice, as a distinct geographical feature, exhibits both aggregated distribution in large ice floes and scattered distribution in small ice fragments within the ocean. Traditional semantic segmentation networks face challenges in accurately extracting sea ice, including imprecise edge extraction and the tendency to overlook fine ice fragments. To address these issues, we propose a novel sea ice extraction model called CA-DeepLabV3+ that incorporates a coordinate attention mechanism. By integrating the coordinate attention module after the ASPP module in DeepLabV3+, our model effectively enhances the feature representation of both channel and spatial dimensions, facilitating the capture of large-scale information and edge details of sea ice. Experimental evaluations are conducted using Sentinel-1 remote sensing imagery collected from the Greenland Sea in the Arctic. A dataset is prepared through remote sensing data processing, annotation, and augmentation. The performance of CA-DeepLabV3+ is compared with that of classical semantic segmentation networks, including UNet, PSPNet, and the original DeepLabV3+. The results demonstrate that CA-DeepLabV3+ achieves an mIoU of 78.10% and an mPA (mean Pixel Accuracy) of 91.16%. Compared to the original DeepLabV3+, the mIoU is improved by 5.27% and the mPA is improved by 1.29%. Moreover, CA-DeepLabV3+ outperforms UNet and PSPNet, achieving the best performance among all the compared networks. It exhibits stronger learning capability for sea ice features, enabling the acquisition of more detailed information, thus providing further technical support for studying sea ice degradation under global warming conditions.

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

Citation Download Citation

Shichang Sun, Zhiyong Wang, and Kang Tian "Fine extraction of arctic sea ice based on CA-DeepLabV3+ model", Proc. SPIE 12797, Second International Conference on Geographic Information and Remote Sensing Technology (GIRST 2023), 127971P (30 August 2023); https://doi.org/10.1117/12.3007422

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

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
6 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Ice

Education and training

Image segmentation

Semantics

Convolution

Deep learning

Synthetic aperture radar

Show All Keywords

Keywords/Phrases

Search In:

Publication Years