BOEM Deep-Sea Nodule Detection

Python • OpenCV • Machine Learning • Streamlit

Overview

This project was part of my work with the Bureau of Ocean Energy Management (BOEM), where I developed an AI-assisted computer vision pipeline to detect deep-sea polymetallic nodules from remotely collected imagery. The goal was to automate early-stage image analysis and support environmental evaluation workflows.

Technical Approach

I built a complete preprocessing and segmentation workflow with OpenCV, including:

Grayscale conversion and adaptive contrast enhancement
Noise reduction and morphological filtering
Dark-spot detection using thresholding + blob analysis
Patch-based extraction for fine-grained nodule classification
Overlay visualization for inspection and annotation

Interactive Visualization Tool

To make the pipeline accessible to collaborators, I created a Streamlit GUI that allows users to:

Upload raw seafloor imagery
Preview preprocessing stages
Adjust thresholds and parameters
Visualize detected nodules with bounding overlays

This made the system usable for team members without coding or computer vision background.

Impact

The tool demonstrated how lightweight AI can meaningfully accelerate environmental image analysis, automate early triage of large datasets, and support BOEM's mission of understanding deep-sea mineral distributions.