Demand-driven and resource-efficient hydrographic surveying requires prior knowledge of the variability of the seabed. This paper presents an approach to obtain this prior knowledge of changes in seabed topography by means of bathymetric data derived from less accurate but high frequency multispectral satellite imagery and a change analysis based on it. This approach is designed to be implemented as a fully automated operational service at the German Federal Maritime and Hydrographic Agency to provide decision support for the operational planning of hydrographic surveying. Spectrally-derived bathymetry is conducted using a convolutional neural network, with a median absolute error of 0.47 m and a RMSE of 0.86 m. Various change detection techniques such as principle component analysis, change vector analysis, least squares tracking and robust median difference are used for change analysis. The results are weighted together with additional current and wave information and summarised into a single change value. Finally, the data is spatially aggregated and converted into an intuitive traffic light scheme that provides a recommended course of action and enables a more targeted hydrographic surveying.
Read full abstract7-days of FREE Audio papers, translation & more with Prime
7-days of FREE Prime access