Source detection and tracking for underwater distributed acoustic sensing

Distributed Optical Fiber Sensing (DOFS) transforms conventional fiber optic cables into an extensive network of continuous sensors. It achieves this by exploiting the spectral, polarization and/or phase sensitivity of the propagating light to measurands of temperature, strain, pressure, vibrations etc. To harness the novel capabilities of optical fibers to remotely capture, process and coherently analyze ambient vibration (e.g., acoustic) fields, it is crucial to address the challenges of the diversity of noise introduced in DOFS measurements, in particular, within the under-explored submarine environment. This research introduces a comprehensive workflow for the detection of active (uncontrolled) acoustic sources, comprised of successive denoising steps that deal with the distinctive properties of such environments. Leveraging the spatio-temporal density of DOFS measurements, we develop a method based on data covariances for the automatic extraction of features in an unsupervised manner, together with additional features introduced to distinguish active source signals from noise. Consequently, this work takes the denoising of underwater DOFS data one step further through the application of a tracking algorithm on real, novel submarine DOFS data, laying the foundation for broader applications of DOFS data analysis in marine environmental sensing and monitoring.