Bathymetry Inversion

Bathymetry inversion aims to reconstruct underwater terrain (bathymetry) from indirect measurements, such as sidescan sonar intensity or surface flow velocities, addressing the high cost and logistical challenges of direct surveys. Current research heavily utilizes deep learning, employing convolutional neural networks and variational autoencoders to create efficient surrogate models for complex physical processes (e.g., shallow water equations) and to perform the inversion itself. These advancements enable faster and more accurate bathymetry estimation, with significant implications for river management, navigation safety, and environmental monitoring.