MAFUNet: water body segmentation algorithm for SAR images combining attention mechanisms and active contour loss

doi:10.11947/j.AGCS.2025.20240235

Abstract

Abstract:

With the increasing development of remote sensing technology, synthetic aperture radar (SAR) imagery has become one of the main methods for water body segmentation. Due to the presence of complex interference, water body segmentation in SAR imagery is still a challenging task. To achieve accurate water body detection, we proposed a multi-level attention fusion UNet (MAFUNet) inspired by the effectiveness of UNet in segmenting small targets with weak edges. First, the spatial attention module (SAM) and channel attention module (CAM) are added to the skip connections between the encoder and decoder parts to extract useful low-level and high-level features, compensating for the loss of semantic information of downsampling. Second, considering the feature distortion resulting from upsampling, an attention upsampler (AU) is designed that retains more detailed information of the image and reduces the effect of introduced noise. Third, the multiscale convolutional pooling block (MCPB) is introduced into the decoder part to better utilize the contextual information, capturing water and land features at different scales. Moreover, an active contour loss is designed as an additional regularization, and a multilayer loss function is used to optimize the network for better extraction of layer-level features, improving the model's segmentation performance. The experimental results show that the proposed MAFUNet outperforms other state-of-the-art models, IoU and F₁ are 94.28% and 97.67% on the ALOS PALSAR dataset.

Key words: water body segmentation, SAR, attention fusion UNet, MCPB, active contour loss

CLC Number:

P234

Guangao XING, Guanming LU, Bin HAN. MAFUNet: water body segmentation algorithm for SAR images combining attention mechanisms and active contour loss[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(5): 924-936.

Figures/Tables 15

Fig. 1

Tab. 1

Fig. 2

Fig. 3

Fig. 4

Tab. 2

Fig. 5

Tab. 3

Fig. 6

Tab. 4

Tab. 5

Tab. 6

Tab. 7

Tab. 8

Fig. 7

References 30

[1]	ZHAO Bofei, SUI Haigang, ZHU Yihao, et al. Real-time rescue target detection based on UAV imagery for flood emergency response[J]. Journal of Geodesy and Geoinformation Science, 2024, 7(1): 74-89.
[2]	韩斌, 吴一全. SAR图像河流提取的主动轮廓模型的稳健估计算法[J]. 测绘学报, 2020, 49(6): 777-786. DOI . doi: 10.11947/j.AGCS.2020.20180423
	HAN Bin, WU Yiquan. Robust estimation algorithm of active contour model for river extraction in SAR images[J]. Acta Geodaetica et Cartographica Sinica, 2020, 49(6): 777-786. DOI . doi: 10.11947/j.AGCS.2020.20180423
[3]	杨蕴, 李玉, 赵泉华. 斑点抑制与多分辨率拓扑分析相结合的SAR图像河流水体提取[J]. 测绘学报, 2022, 51(1): 145-158. DOI . doi: 10.11947/j.AGCS.2022.20190395
	YANG Yun, LI Yu, ZHAO Quanhua. River waterbody extraction from SAR images based on speckle reduction and multi-resolution topological analysis[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(1): 145-158. DOI . doi: 10.11947/j.AGCS.2022.20190395
[4]	吴庆双, 汪明秀, 申茜, 等. Sentinel-2遥感图像的细小水体提取[J]. 遥感学报, 2022, 26(4): 781-794.
	WU Qingshuang, WANG Mingxiu, SHEN Qian, et al. Small water body extraction method based on Sentinel-2 satellite multi-spectral remote sensing image[J]. National Remote Sensing Bulletin, 2022, 26(4): 781-794.
[5]	LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks for semantic segmentation[C]//Proceedings of 2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE, 2015: 3431-3440.
[6]	RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]//Proceedings of 2015 Medical Image Computing and Computer-Assisted Intervention. Cham: Springer International Publishing, 2015: 234-241.
[7]	BADRINARAYANAN V, KENDALL A, CIPOLLA R. SegNet: a deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(12): 2481-2495.
[8]	CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Semantic image segmentation with deep convolutional nets and fully connected CRFs[EB/OL]. [2024-03-06]. https://arxiv.org/abs/1412.7062v4.
[9]	CHEN L C, ZHU Yukun, PAPANDREOU G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Proceedings of 2018 ECCV. Cham: Springer International Publishing, 2018: 833-851.
[10]	KIM M U, OH H, LEE S J, et al. Deep learning based water segmentation using KOMPSAT-5 SAR images[C]//Proceedings of 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS. Brussels: IEEE, 2021: 4055-4058.
[11]	ZHAO Hengshuang, SHI Jianping, QI Xiaojuan, et al. Pyramid scene parsing network[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 6230-6239.
[12]	LI Juanjuan, WANG Chao, XU Lu, et al. Multitemporal water extraction of Dongting Lake and Poyang Lake based on an automatic water extraction and dynamic monitoring framework[J]. Remote Sensing, 2021, 13(5): 865.
[13]	DIRSCHERL M, DIETZ A J, KNEISEL C, et al. A novel method for automated supraglacial lake mapping in Antarctica using Sentinel-1 SAR imagery and deep learning[J]. Remote Sensing, 2021, 13(2): 197.
[14]	ZHANG Jinsong, XING Mengdao, SUN Guangcai, et al. Water body detection in high-resolution SAR images with cascaded fully-convolutional network and variable focal loss[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(1): 316-332.
[15]	REN Yibin, LI Xiaofeng, YANG Xiaofeng, et al. Development of a dual-attention U-Net model for sea ice and open water classification on SAR images[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 4010205.
[16]	胡功明, 杨春成, 徐立, 等. 改进U-Net的遥感图像语义分割方法[J]. 测绘学报, 2023, 52(6): 980-989. DOI . doi: 10.11947/j.AGCS.2023.20210684
	HU Gongming, YANG Chuncheng, XU Li, et al. Improved U-Net remote sensing image semantic segmentation method[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(6): 980-989. DOI . doi: 10.11947/j.AGCS.2023.20210684
[17]	韩斌, 吴一全. SAR图像河流分割的加权指数区域能量模型[J]. 测绘学报, 2017, 46(9): 1174-1181. DOI . doi: 10.11947/j.AGCS.2017.20170134
	HAN Bin, WU Yiquan. Weighted exponential region energy model for river segmentation of SAR images[J]. Acta Geodaetica et Cartographica Sinica, 2017, 46(9): 1174-1181. DOI . doi: 10.11947/j.AGCS.2017.20170134
[18]	TORRALBA A, RUSSELL B C, YUEN J. LabelMe: online image annotation and applications[J]. Proceedings of the IEEE, 2010, 98(8): 1467-1484.
[19]	OKTAY O, SCHLEMPER J, LE FOLGOC L, et al. Attention U-Net: learning where to look for the pancreas[EB/OL]. [2024-03-06]. https://arxiv.org/abs/1804.03999v3.
[20]	YU Jie, CAI Yang, LYU Xin, et al. Boundary-guided semantic context network for water body extraction from remote sensing images[J]. Remote Sensing, 2023, 15(17): 4325.
[21]	XU Ming, LI Hongping, YUN Yuying, et al. End-to-end pixel-wisely detection of oceanic eddy on SAR images with stacked attention network[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 16: 9711-9724.
[22]	DING Lei, ZHENG Kai, LIN Dong, et al. MP-ResNet: multipath residual network for the semantic segmentation of high-resolution PolSAR images[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 4014205.
[23]	YUE Zhenyu, GAO Fei, XIONG Qingxu, et al. A novel attention fully convolutional network method for synthetic aperture radar image segmentation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13: 4585-4598.
[24]	PAN Bin, SHI Xianjian. Fusing ascending and descending time-series SAR images with dual-polarized pixel attention UNet for landslide recognition[J]. Remote Sensing, 2023, 15(23): 5619.
[25]	WANG Ying, PENG Yuexing, LI Wei, et al. DDU-Net: dual-decoder-U-net for road extraction using high-resolution remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-12.
[26]	FAN Xiangsuo, YAN Chuan, FAN Jinlong, et al. Improved U-Net remote sensing classification algorithm fusing attention and multiscale features[J]. Remote Sensing, 2022, 14(15): 3591.
[27]	ZHANG Xinyu, LI Jinjiang, HUA Zhen. MRSE-Net: multiscale residuals and SE-attention network for water body segmentation from satellite images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 5049-5064.
[28]	HU Jie, SHEN Li, SUN Gang. Squeeze-and-excitation networks[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 7132-7141.
[29]	LI Xiang, WANG Wenhai, HU Xiaolin, et al. Selective kernel networks[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 510-519.
[30]	付相为, 单慧琳, 吕宗奎, 等. 基于深度学习的合成孔径雷达图像去噪算法[J]. 光学学报, 2023, 43(6): 0610002.
	FU Xiangwei, SHAN Huilin, LÜ Zongkui, et al. Synthetic aperture radar image denoising algorithm based on deep learning[J]. Acta Optica Sinica, 2023, 43(6): 0610002.

层	输入	输出	结构
Conv 0	3×256×256	64×256×256	（3×3Conv+BN+ReLU）×2
Layer 1	64×256×256	64×128×128	7×7Conv+BN+ReLU
Layer 2	64×128×128	64×64×64
Layer 3	64×64×64	128×32×32
Layer 4	128×32×32	256×16×16
Layer 5	256×16×16	512×8×8

数据集划分	数据集	样本数量	图像尺寸
训练集		1413	256×256
验证集	ALOS PALSAR	706	256×256
测试集		235	256×256

模型	OA	Pre	R	IoU	F₁值
UNet	88.51	87.53	88.52	89.22	88.02
AUNet	88.01	89.09	90.92	91.15	90.01
BSGNet	88.89	88.45	89.56	90.32	89.00
SANet	93.34	92.02	92.35	92.35	92.18
MP-ResNet	93.56	92.78	92.68	92.56	92.45
DAUNet	94.04	93.02	93.13	92.41	93.08
MAFUNet	97.23	97.69	97.66	94.28	97.67

模型	OA	Pre	R	IoU	F₁值
DAUNet	94.04	93.02	93.13	92.41	93.08
AFCNet	93.68	91.78	91.26	91.47	91.51
DPPAUNet	94.23	93.85	92.62	92.22	93.23
DDUNet	93.62	91.21	91.63	90.42	91.41
SAUNet	96.23	94.52	93.26	92.63	93.88
MRSE-Net	96.63	94.25	94.33	93.62	94.29
MAFUNet	97.23	97.69	97.66	94.28	97.67

（α₁，β₁）	IoU	F₁值
（0，1）	91.68	94.23
（0.25，0.75）	92.88	96.43
（0.5，0.5）	92.72	95.73
（0.75，0.25）	92.65	95.56
（1，0）	90.68	93.02