Water-Body Segmentation for SAR Images: Past, Current, and Future

Abstract

Synthetic Aperture Radar (SAR), as a microwave sensor that can sense a target all day or night under all-weather conditions, is of great significance for detecting water resources, such as coastlines, lakes and rivers. This paper reviews literature published in the past 30 years in the field of water body extraction in SAR images, and makes some proposals that the community working with SAR image waterbody extraction should consider. Firstly, this review focuses on the main ideas and characteristics of traditional water body extraction on SAR images, mainly focusing on traditional Machine Learning (ML) methods. Secondly, how Deep Learning (DL) methods are applied and optimized in the task of water-body segmentation for SAR images is summarized from the two levels of pixel and image. We also pay more attention to the most popular networks, such as U-Net and its modified models, and novel networks, such as the Cascaded Fully-Convolutional Network (CFCN) and River-Net. In the end, an in-depth discussion is presented, along with conclusions and future trends, on the limitations and challenges of DL for water-body segmentation.

U-net-based semantic classification for flood extent extraction using SAR imagery and GEE platform: A case study for 2019 central US flooding

“…This supports recent findings in which the authors claimed that CNN-based DL models, such as U-Net and DeepLabV3+ (Verma et al, 2021), failed to accurately extract narrow rivers captured by SAR images . Some previous studies have attributed this issue to inherent speckle fluctuation in SAR images Moharrami et al, 2021) but we believe it is also due to the mechanism of U-Net and other similar CNN-based semantic segmentation DL models, which downsample images several times to extract abstract features (Guo et al, 2022) on images at the expense of reducing image resolution and then restores to the previous resolution by up-sampling. This down-and up-sampling will inevitably lose some details from the original images, result in the smoothing effect seen in Figure 3, and lead to the slight dilatation for narrow water channels.…”

Monitoring Braided River-Bed Dynamics at the Sub-Event Time Scale Using Time Series of Sentinel-1 SAR Imagery

“…Various techniques are widely used in research fields such as medical applications, the recognition and tracking of objects, and environmental analysis, including the delineation of river and waterbodies from optical, multispectral and radar data. Two groups can be mentioned: traditional methods (e.g., edge detection, clustering, random forest, support vector machine, Markov random field, statistical algorithm), and segmentation processes based on the latest Deep Learning (DL) methods (ANN, CNN, and others) [74].…”

HA-Unet: A Modified Unet Based on Hybrid Attention for Urban Water Extraction in SAR Images

“…It enables CNNs to pay attention to specific parts of the input image and select high-value information from massive information. Furthermore, related studies have shown that the attention mechanism is a means for deep learning models to understand SAR images with complex scenes in terms of accuracy and efficiency [14].…”

“…A variety of traditional methods have achieved a great success for suburban water extraction, but the transferability of these methods for urban water extraction is a critical issue [10][11][12][13]. Very recently, deep learning methods, especially Convolutional Neural Networks (CNNs), enabled remarkable performance in water extraction from SAR images by virtue of its powerful feature extraction ability without auxiliary data [14,15]. SAR image segmentation with CNNs is a task that requires the integration of feature maps from different spatial scales and a balance between local information and global information.…”