Author(s): Xing Xuanwei; Xue Yuan; Zhang Yongxian; Qin Chao; Li Dan; Xu Mengzhen
Linked Author(s): Mengzhen Xu
Keywords: Multi-source remote sensing; River surfaces; Riverbed characteristics; Machine learning
Abstract: River surfaces serve as a type of river boundary that can be extensively observed and automatically extracted using remote sensing. Accurate river surfaces provide vital data for the river research, such as research on watershed-scale river evolution, total water resources assessment, and river carbon dioxide emission estimation. However, extracting extremely small river surfaces in arid and semi-arid areas from satellite imagery remains a challenging issue, which can easily lead to missing river information. In this study, we proposed GRF-CNN method for the extraction of river surfaces based on small river characteristics and deep learning, in order to tackle the challenge of extracting greater abundance of river surfaces during dry season. This study, focused on the Wuding River Basin, a primary tributary of the Yellow River Basin, and utilizes multi-source high-resolution remote sensing data to extract entire river surfaces of the whole river basin. This method allows for the acquisition of riverbed widths without the need for complex in-situ measurements. The results showed that the Kappa coefficient of GRF-CNN is 0.92, with a river extraction accuracy of approximately 93.1%. Compared to existing research, this method achieved a 25.6% improvement in accuracy. The study provided methods and data supports for the extraction of river geometric and morphological information in arid and semi-arid areas.
Year: 2024