Machine Learning Algorithms for On-Ship Radar Clutter Mitigation
Abstract
Publicly Releasable NEEC Topic: DD-01 NEEC Technical POC: Karen Smith Understanding the radar signatures of ambient objects (known as “clutter”) and how to differentiate them from more defense-oriented targets has been an important topic in the design of air defense systems for decades. Clutter can consist of objects on the surface, such as mountains or buildings, or in the air, such as precipitation, birds, or insects. This project will conduct research into leveraging modern machine learning techniques to improve capabilities to classify tracks from radar systems as either clutter or non-clutter. The goal will be to produce a system that is capable of being integrated into shipboard data collection systems and robust enough to reduce operator-in-the-loop functions. The research will investigate both supervised and unsupervised learning approaches for object classification. It will also evaluate the application of strategies such as active learning to supplement existing labeled datasets. Finally, the research will consider strategies for characterizing model uncertainty and improving model robustness in operational scenarios. The result will provide an important progress toward autonomous detection capabilities to improve strategic awareness and defense
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 11, 2025
- Source ID
- N001742310019
Entities
People
- Justin Krometis
Organizations
- United States Navy
- Virginia Tech