article · IEEE Access
Segmenting river water accurately from close-range remote sensing imagery is difficult because water surfaces reflect varying colours and textures from skies and riverbank structures. An evaluation comparing multiple deep learning models across four river scene datasets reveals distinct trade-offs between precision and speed. The fine-tuned Segment Anything Model achieves the highest accuracy, closely followed by U-Net with a ResNet50 backbone, although both require greater computational resources. Conversely, PSPNet with a ResNet50 backbone delivers lower accuracy but provides the fastest execution times. Alongside these performance comparisons, a new benchmarking dataset, LuFI-RiverSnap.v1, provides diverse river scenes with accurate segmentation masks. Fine-tuned model implementations offer tools to track changes in river courses, analyse water level trends, examine river ecosystems, and assist in disaster risk reduction, including rapid flood inundation assessments and automated watershed gauge data extraction.
Monitoring rivers through remote sensing is critical for managing freshwater ecosystems and responding to flood emergencies. However, shifting reflections and riverbank textures often mislead automated image analysis. Identifying the most effective machine learning models and providing higher-quality training datasets helps environmental agencies and emergency services assess flood waters rapidly, track water level changes reliably, and protect surrounding communities.
This applied and tested research provides practical tools for environmental monitoring bodies, disaster response agencies, and watershed managers. The fine-tuned models and dataset can support risk assessment platforms and automated flood monitoring systems by enabling rapid flood inundation mapping and watershed gauge extraction. Because these models have been validated across datasets but still require integration into operational pipelines, the technology sits at an applied, pre-commercial stage.
AI-generated from the published abstract. Always read the original work before citing.
Accurate segmentation of river water in close-range Remote Sensing (RS) images is vital for efficient environmental monitoring and management. However, this task poses significant difficulties due to the dynamic nature of water, which exhibits varying colors and textures reflecting the sky and surrounding structures along the riverbanks. This study addresses these complexities by evaluating and comparing several well-known deep-learning (DL) techniques on four river scene datasets. To achieve this, we fine-tuned the recently introduced "Segment Anything Model" (SAM) along with popular DL segmentation models such as U-Net, DeepLabV3+, LinkNet, PSPNet, and PAN, all using ResNet50 pre-trained on ImageNet as a backbone. Experimental results highlight the diverse performances of these models in river water segmentation. Notably, fine-tuned SAM demonstrates superior performance, followed by U-Net <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">(ResNet50)</sub> , despite their higher computational costs. In contrast, PSPNet <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">(ResNet50)</sub> , while less effective, proves to be the most efficient in terms of execution time. In addition to these findings, we introduce a novel river water segmentation dataset, LuFI-RiverSnap. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">v</i> 1 (Dataset link), characterized by a more diverse range of scenes and accurate masks compared to existing datasets. To facilitate reproducible research in remote sensing and computer vision, we release the implementations of the fine-tuned SAM model (Code link). The findings from this research, coupled with the presented dataset and the accuracy achieved by fine-tuned SAM segmentation, can support tracking river changes, understanding river water level trends, and exploring river ecosystem dynamics. These can also provide valuable insights for practitioners and researchers seeking models tailored to specific image characteristics with practical means in disaster risk reduction, such as rapid assessments of inundations during floods or automatic extractions of gauge data in watersheds.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.1109/access.2024.3385425
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
