Multi-level Feature Attention Fusion Network for Water Extraction from Multi-source SAR Images

CHEN Lifu; LONG Fengqi; LI Zhenhong; YUAN Zhihui; ZHU Wu; CAI Xingmin

doi:10.13203/j.whugis20230041

您当前的位置：

首页 >

文章列表页 >

Multi-level Feature Attention Fusion Network for Water Extraction from Multi-source SAR Images

Photogrammetry and Remote Sensing | 更新时间：2026-11-05

- Multi-level Feature Attention Fusion Network for Water Extraction from Multi-source SAR Images
- Geomatics and Information Science of Wuhan University Vol. 50, Issue 7, Pages: 1339-1345(2025)
- 作者机构：
  
  1.长沙理工大学电气与信息工程学院，湖南长沙，410004
  2.长安大学地质工程与测绘学院，陕西西安，710054
  3.长安大学地学与卫星大数据研究中心，陕西西安，710054
  4.自然资源部生态地质与灾害防控重点实验室，陕西西安，710054
- 作者简介：
- 基金信息：
- DOI：10.13203/j.whugis20230041
  CLC： P237
- Received：20 April 2023，
  
  Published：05 July 2025
- 稿件说明：
移动端阅览
陈立福,龙凤琪,李振洪，等.面向多源SAR图像的多级特征注意力水体提取网络[J].武汉大学学报（信息科学版）,2025,50(7):1339-1345.

CHEN Lifu,LONG Fengqi,LI Zhenhong,et al.Multi-level Feature Attention Fusion Network for Water Extraction from Multi-source SAR Images[J].Geomatics and Information Science of Wuhan University,2025,50(7):1339-1345.
陈立福,龙凤琪,李振洪，等.面向多源SAR图像的多级特征注意力水体提取网络[J].武汉大学学报（信息科学版）,2025,50(7):1339-1345. DOI：10.13203/j.whugis20230041

CHEN Lifu,LONG Fengqi,LI Zhenhong,et al.Multi-level Feature Attention Fusion Network for Water Extraction from Multi-source SAR Images[J].Geomatics and Information Science of Wuhan University,2025,50(7):1339-1345. DOI：10.13203/j.whugis20230041

摘要

利用合成孔径雷达（synthetic aperture radar，SAR）图像进行水体提取在城市水体监测、海岸线监测、洪涝灾害监测等方面具有重要的应用价值。目前对单频段SAR图像数据水体提取已经取得了较大成功。但由于水体及地物目标在不同频段SAR图像中呈现的特性有较大差别，针对单频段SAR图像设计的深度学习网络，在应用到不同频段时提取精度较差，因此，如何实现多源SAR图像水体自动提取依然是个不小的挑战。对此，提出了一种新的网络框架，即多级特征注意力融合网络（multi-level feature attention fusion network，MFAFNet)。该网络由编码器和解码器组成，编码器先使用ResNet-101提取具有不同分辨率的4级特征，再由所提出的中间级特征融合模块和有效通道空洞空间卷积池化金字塔模块并行处理，对中高级特征进行深度融合；在解码器中引入注意调制模块对低级特征进行权重分配，进一步与来自编码器的高级特征融合并进一步处理，获得水体提取结果。为了验证所提网络框架的有效性，对不同频段和分辨率的SAR图像（哨兵1号、TerraSAR和高分三号）进行了实验，并与3个典型网络进行了对比。结果表明，MFAFNet对多源SAR图像的水体提取效果显著优于其他网络，平均用户精度可达87%，平均交并比为0.80，实现了多源SAR图像水体的高精度自动提取。

Abstract

Objective

At present

water extraction from single frequency-band synthetic aperture radar （SAR） images has achieved great success. However

due to the large differences in the characteristics of water bodies and other surface targets in SAR images with different frequency-bands

deep learning network designed for single frequency-band has poor extraction accuracy when applied to SAR images with different frequency bands. So

how to achieve excellent extraction performance for water automatically from multi-source SAR images is still a challenge.

Methods

We propose a new network framework

namely multi-level feature attention fusion network (MFAFNet). The network consists of encoder and decoder. The encoder utilizes ResNet-101 to generate four-level features with different resolutions

which are then processed in parallel by the intermediate level feature fusion module and efficient channel atrous spatial pyramid pooling module

and the intermediate and high-level features are output after deep fusion. The attention modulation module attention mechanism is introduced into decoder to distribute the weight of low-level features

then which are further fused with the high-level features from the encoder to obtain the water extraction results.

Results

Experiments are carried out on SAR images with different frequency bands and resolutions (Sentinel-1

TerraSAR and Gaofen-3)

and three good networks have been compared. The results show that the MFAFNet extraction effect is obviously better than other networks

with an average pixel accuracy of 87% and an average intersection over union of 0.80.

Conclusions

The water body extraction from multi-band and multi-resolution SAR images is achieved

which promotes the practical application value of deep learning network in water body detection.

关键词

Keywords

references

MARTINIS S ， GROTH S ， WIELAND M ， et al . Towards a Global Seasonal and Permanent Reference Water Product from Sentinel-1/2 Data for Improved Flood Mapping ［J］. Remote Sensing of Environment ， 2022 ， 278 ： 113077 .

CHEN L F ， ZHANG P ， XING J ， et al . A Multi-scale Deep Neural Network for Water Detection from SAR Images in the Mountainous Areas ［J］. Remote Sensing ， 2020 ， 12 （ 19 ）： 3205 .

ZHANG P ， CHEN L F ， XING J ， et al . Water and Shadow Extraction in SAR Image Based on a New Deep Learning Network ［J］. Sensors ， 2019 ， 19 （ 6 ）： 3576 .

XIE C ， HUANG X ， ZENG W X ， et al . A Novel Water Index for Urban High-Resolution Eight-Band WorldView-2 Imagery ［J］. International Journal of Digital Earth ， 2016 ， 9 （ 10 ）： 925 - 941 .

YAO F F ， WANG C ， DONG D ， et al . High-Resolution Mapping of Urban Surface Water Using ZY-3 Multi-spectral Imagery ［J］. Remote Sensing ， 2015 ， 7 （ 9 ）： 12336 - 12355 .

LECUN Y ， BENGIO Y ， HINTON G . Deep Learning ［J］. Nature ， 2015 ， 521 （ 7553 ）： 436 - 444 .

LI Z Y ， WANG R ， ZHANG W ， et al . Multiscale Features Supported DeepLabV3+ Optimization Scheme for Accurate Water Semantic Segmentation ［J］. IEEE Access ， 2019 ， 7 ： 155787 - 155804 .

DUAN L H ， HU X Y . Multiscale Refinement Network for Water-Body Segmentation in High-Resolution Satellite Imagery ［J］. IEEE Geoscience and Remote Sensing Letters ， 2020 ， 17 （ 4 ）： 686 - 690 .

LI Y S ， DANG B ， ZHANG Y J ， et al . Water Body Classification from High-Resolution Optical Remote Sensing Imagery： Achievements and Perspectives ［J］. ISPRS Journal of Photogrammetry and Remote Sensing ， 2022 ， 187 ： 306 - 327 .

CHU Z Q ， TIAN T ， FENG R Y ， et al . Sea-Land Segmentation with ResUNet and Fully Connected CRF ［C］// IEEE International Geoscience and Remote Sensing Symposium ， Yokohama， Japan ， 2019 .

覃星力，杨杰，李平湘，等 . 迁移学习用于多时相极化SAR影像的水体提取［J］. 武汉大学学报（信息科学版）， 2022 ， 47 （ 7 ）： 1093 - 1102 .

QIN Xingli ， YANG Jie ， LI Pingxiang ， et al . Water Body Extraction from Multi-temporal Polarimetric SAR Images Based on Transfer Learning ［J］. Geomatics and Information Science of Wuhan University ， 2022 ， 47 （ 7 ）： 1093 - 1102 .

眭海刚，陈光，胡传文，等 . 光学遥感影像与GIS数据一体化的水体分割、配准与提取方法［J］. 武汉大学学报（信息科学版）， 2016 ， 41 （ 9 ）： 1145 - 1150 .

SUI Haigang ， CHEN Guang ， HU Chuanwen ， et al . Integrated Segmentation， Registration and Extraction Method for Water-Body Using Optical Remote Sensing Images and GIS Data ［J］. Geomatics and Information Science of Wuhan University ， 2016 ， 41 （ 9 ）： 1145 - 1150 .

HE K M ， ZHANG X Y ， REN S Q ， et al . Deep Residual Learning for Image Recognition ［C］// IEEE Conference on Computer Vision and Pattern Recognition ， Las Vegas， USA ， 2016 .

WANG Q L ， WU B G ， ZHU P F ， et al . ECA-Net： Efficient Channel Attention for Deep Convolutional Neural Networks ［C］//IEEE/CVF Conferen‑ce on Computer Vision and Pattern Recognition， Seattle， WA， USA ， 2020 .

CAI X M ， CHEN L F ， XING J ， et al . Automatic Extraction of Layover from InSAR Imagery Based on Multilayer Feature Fusion Attention Mechanism ［J］. IEEE Geoscience and Remote Sensing Letters ， 2021 ， 19 ： 4017705 .

CHEN L F ， LUO R ， XING J ， et al . Geospatial Transformer Is What You Need for Aircraft Detection in SAR Imagery ［J］. IEEE Transactions on Geoscience and Remote Sensing ， 2022 ， 60 ： 5225715 .

LUO R ， CHEN L F ， XING J ， et al . A Fast Aircraft Detection Method for SAR Images Based on Efficient Bidirectional Path Aggregated Attention Network ［J］. Remote Sensing ， 2021 ， 13 （ 15 ）： 2940 .

QIN J ， WU J ， XIAO X F ， et al . Activation Modulation and Recalibration Scheme for Weakly Supervised Semantic Segmentation ［J］. Proceedings of the AAAI Conference on Artificial Intelligence ， 2022 ， 36 （ 2 ）： 2117 - 2125 .

CHEN L C ， ZHU Y K ， PAPANDREOU G ， et al . Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation ［C］// The European Conference on Computer Vision ， Munich， Germany ， 2018 .

XIE E Z ， WANG W H ， YU Z D ， et al . SegFormer： Simple and Efficient Design for Semantic Segmentation with Transformers ［J］. Advances in Neural Information Processing Systems ， 2021 ， 34 ： 12077 - 12090 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Water Extraction of High Spatial Resolution Remote Sensing Image Fusing Scale Attention Mechanism

DCMR-Net： A Multi-scale Residual Denoising Network for Seismic Records with Complex Noise

Point⁃Annotation⁃Based Weakly Supervised Semantic Segmentation of Multi⁃source Remote Sensing Images via Detail⁃Global Feature Fusion and Structure Enhancement

Advances in Multimodal Remote Sensing Image Matching： A Survey from Conventional Approaches to Deep Learning

Reservoir Bank Slope Displacement Prediction Model Based on Lagged Attenuation Effect and Key Factor Selection Strategy

Related Author

YU Xin

WANG Xin

ZHENG Zhaobao

MENG Lingkui

LI Linyi

YIN Xin

XIAO Zhou

CHEN Guanzhi

Related Institution

Beijing Institute of Petrochemical Technology

School of Remote Sensing and Information Engineering, Wuhan University

Beijing Institute of Water Resources Planning, Design and Research

State Key Laboratory of Precision Geodesy, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences

School of Earth and Space Sciences, University of Science and Technology of China

AI问答

⁰