Artificial Intelligence tools in the analysis of large seafloor data sets
Abstract
This proposal seeks to address scientific issues and data ultimately related to the Office of Naval researchs Task Force Ocean. The data to support the research was collected over a period of a decade in the global ocean and, particularly, the Pacific. The instruments used in this work were collected by ocean bottom seismographs in shallow ( 100m) and deep water ( 6km) for extended periods (a year). The sensors used for this work included both hydrophones (pressure gauges) and accelerometers. The instrumentation was designed to have very low noise levels across a very broad band of frequencies extending from tides to 50Hz. While traditional data analyses have been conducted on these data for individual instruments, the use of many instruments at common times has been beyond the scope of traditional methods. We intend to exploit new methods in Deep Learning or Artificial Intelligence (AI), to analyze combinations of instruments to understand better phenomena that include tidal noise and signals generated by ships. In many cases, commercial ship positions are available globally through the AIS system. Archives of these data can be used to check the accuracy of the seafloor data that we have collected. Deep learning networks have been used for image classification for many years, they are powerful tools for signal data as well. A deep learning network can do everything a mathematical or physics model can do without requiring the scientist to know which signal features should be brought into the physical model. The network itself decides these features during a learning process.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2020
- Source ID
- N000142012846
Entities
People
- John A. Orcutt
Organizations
- Office of Naval Research
- United States Navy
- University of California, San Diego