FURI | Spring 2025

Occlusion-induced Effects on Sonar Localization

Occlusion-induced effects create significant challenges in sonar detection and imaging, especially in underwater environments where natural and man-made structures obstruct the acoustic wave paths. Non-line-of-sight (NLoS) imaging using sonar allows for the detection of objects obscured by occlusions, such as seafloor topography, underwater structures, and debris. This project aims to create a Convolutional Neural Network to verify that detecting targets around occlusions is possible. The network is trained on spectrograms taken from the NLoS scans to classify which location the obstructing objects lie on our coordinate system. Potential applications include ranging, localization, tracking, and potentially image reconstruction.