Scale-Consistent Multispectral Reconstruction From UAV and Satellite Observations for Coastal Bathymetric Inversion
Abstract
High-resolution quantitative remote sensing is constrained by the spatial–spectral trade-off among current observation platforms: uncrewed aerial vehicle (UAV) imagery provides fine spatial detail but limited spectral dimensionality, whereas satellite multispectral data provide richer spectral information at coarser spatial scales. This study proposes data-driven spectral-structural anchoring fusion (DSSAF), a scale-consistent framework that reconstructs UAV-scale multispectral imagery by integrating fine-resolution structural information from UAV RGB imagery with multispectral constraints from PlanetScope observations. The framework combines land-cover-conditioned spectral mapping, data-driven structural reference selection, adaptive detail injection, and scale-consistency correction. It was evaluated over a heterogeneous intertidal sandbar system and further tested through shallow-water bathymetric inversion. The reconstructed imagery preserved UAV-scale textures, gradients, and edges while remaining radiometrically consistent with the original PlanetScope data after reaggregation to the native satellite scale. In the downstream bathymetry task, DSSAF-fused imagery achieved the best performance among the tested inputs, with R ${}^{2} =0.924$ and RMSE = 0.079 m. These results demonstrate that scale-consistent UAV-satellite fusion can improve both image-level spatial–spectral representation and quantitative geophysical retrieval, providing a transferable approach for high-resolution environmental mapping in heterogeneous coastal landscapes.