ohw26_proj_BoatPhone

Folder Structure

• contributor_folders Each contributor has their own folder and pushes their work here during the week, preventing merge conflicts.
• final_notebooks Final versions of the team's project notebooks live here.
• scripts Shared scripts or functions are added here.
• data Shared datasets are shared here.

Start here

final_notebooks/acoustic_pipeline.ipynb is the reproducible record of the hydrophone (acoustic) pipeline, start to finish: getting an ONC token, writing the .env file, downloading the data, computing band levels, counting vessel passages, and producing every figure. It is written for somebody who has never seen this repository, and it states plainly what the pipeline does and does not establish. A companion notebook for the optical pipeline is in progress; the acoustic notebook flags every point where the two meet.

Project Name

Satellite-derived vessel distance, speed, and orientation relative to the Folger Deep hydrophone array improve the estimation of acoustic ranging for small-scale vessel noise.

One-line Description

This project aims to improve the estimation of small-scale vessel noise detection ranging for the Folger Deep hydrophone array by calibrating acoustic profiles to optically measured distance, speed, and orientation from satellite-derived small-scale vessel detections.

Collaborators

Name	Role
Neve Foreman	Vessel detection model development
Malachy McCaffrey	Optical imagery pipeline development
Isaac Guld	Acoustic data pipeline development

Planning

• Initial idea: Cross-calibrating optical vessel detection against passive acoustics at Folger Deep
• Ideation Slide: https://docs.google.com/presentation/d/1_KLEDpLLvtKpH3awDlZRAiOKuHzbEti4CWmhEykuCG8/edit?slide=id.g3f85357d4e2_21_0#slide=id.g3f85357d4e2_21_0
• Slack channel: ohw26_proj_BoatPhone
• Final presentation: https://docs.google.com/presentation/d/1LnokajlD4dj0683nHa6zb9H3QxUS-2aq0sLbgzFS5yQ/edit?usp=sharing

Background

Satellites see every boat but only for a fraction of a second per day; the hydrophone listens continuously but can't count. Use each to fix the other's blind spot.

Goals

1. Search for, screen, and select PlanetScope imagery scenes for the vessel detection model.
2. Augment open-source machine learning vessel detection model to process selected PlanetScope scenes.
3. Train an ML model or apply signal filtering methods to the hydrophone data to identify acoustic vessel noise signatures, by comparing the hydrophone data to the labeled satelite images

• Create a model that uses hydrophone data and prior knowledge of vessels detected from optical satellite imagery to create continuous estimates of the number of boats, and potential distinguish between larger or smaller boats (those without AIS)

Datasets

• Acoustic data from Folger Deep hydrophone array
• PlanetScope 3m optical imagery

Workflow/Roadmap

Results/Findings

Lessons Learned

References