Automated post-processing, using machine-learning models; Automated analytical tools (that generate estimates of fish frequency, abundance, and distribution)

The Pathway Program - Data Automation: Echosounders

Project: Improving Automated Post-Processing of Echosounder Data with Machine Learning Models.
The developed software, called Echofilter, includes machine-learning models that when used in conjunction with Echoview (Echoview Software Pty Ltd, Hobart, Australia) reduces the post-processing time of hydroacoustic data. Echofilter automates the placement of the full suite of lines defining the boundaries at physical interfaces: the echosounder nearfield, the seafloor, the air-sea interface, and the boundary that defines the extent of penetration of entrained air into the sea. The major contribution of Echofilter is the pronounced, consistent, substantial, and measurable improvement of the automated placement of the entrained air line. The results of a time test demonstrate that implementing the models produced a 45% - 59% reduction in time required to finalize post-processing edits.

Project: Automated Analytical and Reporting Tool.
An automated analytical and reporting tool is being developed in order to automate the production of figures and tables for reporting the frequency, abundance, and distribution of fish targets as recorded in echosounder data. The tool, written in R and R Markdown, uses data exported from Echoview (Echoview Software Pty Ltd, Hobart, Australia) and automates and standardizes the production of quarterly and annual reports. The tool is being developed using WBAT data collected in Minas Passage in 2018.

Team

Principal Investigators: Dr. Scott Lowe, DeepSense & Dr. Louise McGarry

Date
April 1, 2020 – May 31, 2021
Tags