A detection of channel bathymetry from drones: accuracy of point clouds and RGB images for reconstruction of channel-bed topography

Abstract

Channel bathymetry detection focuses on extracting riverbed topography, and it is a primary tool for analysing channel dynamics, sedimentary processes, and habitat structures. A key challenge in underwater bathymetry is accurately capturing high-resolution topography of the channel bed.  Knowledge of underwater channel shape and topography is a fundamental parameter essential for hydraulic modelling, monitoring of channel bed changes, and environmental assessments of riparian ecosystems, including underwater vegetation, grain size, and fish habitats. The objective of this study is to reconstruct the channel topography of the shallow, multi-channel braided Belá River and the artificial Gabčíkovo-Topoľníky canal, featuring shallower and deeper sections, to critically evaluate the efficiency of various UAV-based bathymetric methods. Riverbed topography was reconstructed by two main approaches: through-water photogrammetry and spectral depth estimation. Results were validated through a traditional field survey, which included field measurements and a sonar survey. The accuracy of bathymetry refraction correction achieved a mean error of less than 1 cm, and the root mean square error (RMSE) reached a maximum of 13 cm for the Belá River site, as validated by field depth measurements. The artificial canal Gabčíkovo-Topoľníky exhibited errors (RMSE) ranging from 12 to 30 cm, as validated by sonar. The optical bathymetry approach pointed to errors ranging from 14.7 to 23.3 cm with a regression model R2 from 0.47 to 0.53.