A novel approach for utilizing UAV imaging and deep learning techniques for river channel measurement and flood simulation
Journal
Geomorphology
Journal Volume
496
Start Page
110173
ISSN
0169555X
Date Issued
2026-03-01
Author(s)
Abstract
River hydraulic analysis is a crucial prerequisite for disaster prevention. Traditionally, essential parameters for such analysis are acquired through field observations of the river channel, which are labor and cost intensive. This study employs image sensing techniques and unmanned aerial vehicles (UAVs) to collect extensive spatial information on river channels and generate both two-dimensional and three-dimensional data for rapid updating of riverbed attributes. The gravel size distribution of the riverbed can also be automatically detected and analyzed by deep learning techniques. Gravel parameters in the range of D50−D90 were computed to further estimate the Manning's roughness coefficients. Finally, a hydraulics analysis was performed using the above derived parameters for river flood levels and fluid dynamics simulations. The results demonstrated accuracy of 95.39% in water level estimation as compared with true values. Therefore, the proposed innovative approach can greatly increase the efficiency of river hydraulic analysis and flooding simulation.
Subjects
Deep learning
Flood simulation
Hydraulic analysis
Object detection
UAV
Publisher
Elsevier B.V.
Type
journal article