Machine Learning-Based Tomography in Ocean Acoustics

Abstract

The objective is developing machine learning-based tomography techniques for imaging ocean structures using noise. Specifically, th"e proposed research seeks to (1) further develop a dictionary learning-based travel time tomography approach, accounting for uncertainty in the measurements and physics, (2) formulate the dictionary learning-based approach as a convolutional neuralnetwork (CNN) via convolutional sparse coding (CSC), and (3) apply this CNN tomography framework to ocean acoustic data as a data assimilation framework, to obtain higher-resolution estimates of water column parameters over conventional methods. We will develop (4) sequential Bayesian approaches to geoacoustic inversion for sediment characterization and (5) automated acoustic event detection using machine learning. The proposed research will cover a post doctoral reasercher (Michael J. Bianco, PhD October 2018) and one graduate student ("Dylan Snover).

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 20, 2019
Source ID
N000141912575

Entities

People

  • Peter Gerstoft

Organizations

  • Office of Naval Research
  • United States Navy
  • University of California, San Diego

Tags

Readers

  • Neural Network Machine Learning.
  • Oceanography.
  • Wave Propagation and Nonlinear Chaotic Dynamics.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks