Thesis Work - Seeing Beneath the Surface: A Data Fusion Approach to Bathymetric Mapping

The project will investigate how machine learning / deep learning methods can be used to align, calibrate and merge these heterogeneous, multi-resolution data sources into a single consistent depth model, using the sparse but accurate ROV measurements to correct and validate the wide-area satellite- and drone-derived depth estimates. A further purpose of the thesis is to combine the fused bathymetry with above-water elevation and terrain data to build a full 3D reconstruction of the basin, which will be used to estimate the water volume of the area.

The following steps are foreseen to execute the project:

Literature review of satellite-derived bathymetry, drone-based photogrammetry/LiDAR and ROV sonar surveying methods, and existing data fusion techniques.
Data acquisition and collection of satellite, drone and ROV data for the area of interest.
Pre-process and co-register the multi-source data (satellite, drone, ROV) to a common spatial and depth reference.
Develop and apply a data fusion model (e.g. machine learning / deep learning based) to combine the multi-source bathymetric data into a unified depth map.
Perform elevation and terrain analysis and build a 3D reconstruction of the basin (bathymetry + surrounding terrain) to estimate the water volume.
Evaluate and validate the fused bathymetry model performance against reference/ground-truth measurements. Several field campaigns will be carried out during the thesis period to collect validation data.