2025 CSDMS meeting-085
From CSDMS
(if you haven't already)
Log in (or create account for non-CSDMS members)
Forgot username? Search or email:CSDMSweb@colorado.edu
Browse abstracts
Customizing River Corridor Segmentation from Satellite Imagery Using Deep Learning
Qiuyang Chen,
University of Edinburgh Edinburgh , United Kingdom. qiuyangschen@gmail.com
Simon M. Mudd, University of Edinburgh Edinburgh , United Kingdom.
Mikael Attal, University of Edinburgh Edinburgh , United Kingdom.
Chenyu Zhang, Manchester , United Kingdom.
Mapping river corridors remains challenging due to the dynamic interactions between water, sediment, and vegetation. Existing land cover maps often misclassify fluvial sediments, limiting their use in river system studies. We present a deep learning framework using incremental learning to refine river corridor mapping by integrating Sentinel-2 imagery with global land cover datasets (ESRI, Google Dynamic World, ESA WorldCover). Our method builds on existing classifications to improve differentiation between fluvial sediment, bare ground, and mining-related disturbances. The results show that incremental learning can enhance river mapping accuracy, providing a customizable approach to better capture riverine landscapes.
