Optically derived 3D models for munitions location and identification
Budget UdG: $270.8 kUSD
Duration: 05/2024 – 12/2026
PI: Nuno Gracias
The University of Girona is a key partner in a research project aimed at optimizing the use of optical data to locate and identify underwater military munitions (UWMM). The project seeks to quantify the optimal strategies and limitations of optical sensing for UWMM detection by analyzing critical factors such as resolution, noise, visibility, biofouling, and burial conditions.
Focusing on the intersection of optical sensors and machine learning, the project follows three main objectives:
Dataset Acquisition – Capturing a comprehensive underwater imagery dataset of UWMM across various environmental conditions, substrates, and burial states.
Data Quality Analysis – Evaluating how environmental and imaging parameters impact optical data quality.
Machine Learning Applications – Addressing data fusion challenges and limited training data issues for UWMM classification.
The project involves assembling a suite of 40-50 inert munitions and an equal number of other man-made objects (“clutter”), with half exposed to seawater over three years to induce natural corrosion and biofouling. These objects will be imaged at regular intervals across different underwater environments, including sandy, seagrass, rocky, and mixed habitats, ensuring a diverse dataset for training classification models. High-resolution 3D models generated through photogrammetry will be used to evaluate advanced automated classification methods.
As a project partner, UdG is responsible for developing machine learning architectures designed to work with limited training data. Additionally, the university will create methods for generating realistic synthetic training data to enhance machine learning model performance.
This research will significantly advance the use of optical sensing and AI-driven classification for underwater munitions detection, contributing to safer and more effective survey methodologies.
Partners:
University of Miami
White River Technologies, Inc.
University of Girona