| 引用本文: | 张春芳,刘 鹏,王瑞丽,等.基于可变形卷积视觉模型的极化SAR数据洪水检测方法研究[J].灌溉排水学报,2025,44(9):133-142. |
| ZHANG Chunfang,LIU Peng,WANG Ruili,et al.基于可变形卷积视觉模型的极化SAR数据洪水检测方法研究[J].灌溉排水学报,2025,44(9):133-142. |
|
| 摘要: |
| 【目的】针对现有SAR影像洪水检测模型精度不足及专用数据集匮乏的问题，构建基于极化SAR数据的洪水范围检测数据集，并提出一种融合可变形卷积（Deformable Convolutional Network v3，DCNv3）与视觉变压器（Vision Transformer，ViT）的新型深度学习网络模型FWSARNet，以提升SAR影像洪水范围检测的精度和鲁棒性。【方法】首先利用Sentinel-1卫星的极化SAR影像构建了洪水范围检测数据集，并进行了数据增强处理。其次，设计了一种融合可变形卷积（DCNv3）和视觉变压器（ViT）的高效特征提取模块。基于此模块，构建了名为FWSARNet的深度学习网络模型，用于SAR影像的洪水范围检测，并通过所建数据集对模型进行训练与验证。【结果】所提出的FWSARNet模型对复杂地表具有较强的适应性。与其他深度学习网络模型相比，FWSARNet模型在水体边缘、小面积水体及细长线状水体的识别上表现更优，精度更高。在自建的Henan720和Hebei727洪水范围检测数据集上，该模型的平均交并比（mIoU）分别达到了88.53%和92.50%。【结论】构建的模型洪水范围提取精度能够满足应急救灾需求，适用于极化SAR数据处理。 |
| 关键词: 洪水范围检测；极化SAR；可变形卷积；视觉变压器 |
| DOI：10.13522/j.cnki.ggps.2025008 |
| 分类号: |
| 基金项目: |
|
| A deep learning approach for flood inundation mapping in polarimetric SAR images using DCNv3 and vision transformer |
|
ZHANG Chunfang, LIU Peng, WANG Ruili, PAN Deng, MENG Wenmin, YU Haiyang
|
|
1. Henan Jiaozuo Hydrology and Water Resources Forecasting Sub-center, Jiaozuo 454003, China;
2. Land and Spatial Survey and Planning Institute of Henan Province, Zhengzhou 450053, China; 3. Key Laboratory of Mine Spatio-temporal Information and Ecological Restoration, Ministry of Natural Resources, Henan Polytechnic University, Jiaozuo 454003, China;
4. Institute of Natural Resources Monitoring and Comprehensive Land Improvement of Henan Province, Zhengzhou 450016, China
|
| Abstract: |
| 【Objective】Accurate flood inundation detection using Synthetic Aperture Radar (SAR) images remains challenging due to limitations in existing models and the lack of high-quality annotated datasets. This study aims to address these issues by developing a dedicated flood inundation detection dataset based on polarimetric SAR data and proposing a novel deep learning model, FWSARNet, that integrates Deformable Convolutional Networks v3 (DCNv3) and Vision Transformer (ViT) to improve detection accuracy and robustness.【Method】A polarimetric SAR-based dataset was constructed using Sentinel-1 imagery, with extensive data augmentation to enhance model generalization. An efficient feature extraction module was designed by combining DCNv3’s spatial adaptability with ViT’s global feature modeling. This module served as the backbone of the FWSARNet model, which was then trained and validated on two custom-built datasets: Henan720 and Hebei727.【Result】The proposed FWSARNet model outperformed existing deep learning models in delineating complex flood features, including water body edges, small patches, and narrow linear segments. It achieved mean Intersection over Union (mIoU) values of 88.53% on Henan720 and 92.50% on Hebei727, indicating superior performance in diverse flood scenarios.【Conclusion】FWSARNet demonstrates high accuracy and adaptability in flood inundation detection from SAR images and is well-suited for emergency disaster response applications using polarimetric SAR data. |
| Key words: lood extent detection; polarimetric SAR; deformable convolution; vision transformer |
