Bathymetry Estimation using Machine Learning Techniques from Airborne Remote Sensing RGB Image Data

The intertidal zones of the Wadden Sea and estuaries are highly dynamic natural environments subject to constant spatial and temporal fluctuations. This study presents an innovative method for bathymetry estimation based on airborne RGB remote sensing imagery. Traditional acoustic in-situ measurements, while highly accurate, are limited by high costs and restricted accessibility. Our approach investigates color gradients and wave structures of the water surface using a k-means classifier to correlate optical properties with water depths. Additionally, surface waves are analyzed through two-dimensional Fourier transformation to extract wavelength and orientation. Initial results demonstrate a strong correlation between classified image data and reference bathymetric data, highlighting the suitability of this method for rapid and large-scale mapping of dynamic coastal areas. This technique provides a complementary tool to conventional methods, effectively capturing and documenting rapid changes such as sediment shifts and sandbank formation.