Neptune: An Automated System for Dark Ship Detection, Targeting, and Prioritization

The ability to detect dark ships at open-ocean scale requires enhanced space-based intelligence, surveillance, and reconnaissance capabilities. With the boom of commercial space-based sensing, the nation needs an automated process to meet the growing volume and velocity of data. Multimodal data from...

Full description

Saved in:
Bibliographic Details
Published in:Johns Hopkins APL technical digest 2022-01, Vol.36 (2), p.82
Main Authors: Byerly, Adam B, Zhang, William C, Iwarere, Sesan A, Malik, Waseem A, Bish, Sheldon F, Haque, Musad A, Sookoor, Tamim I
Format: Article
Language:English
Subjects:
Anomalies
Automation
Feature extraction
Infrared radar
Microprocessors
Pattern analysis
Ships
Space flight
Synthetic aperture radar
Target recognition
Tracking
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The ability to detect dark ships at open-ocean scale requires enhanced space-based intelligence, surveillance, and reconnaissance capabilities. With the boom of commercial space-based sensing, the nation needs an automated process to meet the growing volume and velocity of data. Multimodal data from the variety of existing and proposed space-based sensor networks can be aggregated and fused to produce target-quality tracks on ships. These sensor modalities include synthetic aperture radar (SAR), electro-optical/infrared (EO/IR), and Automatic Identification System (AIS). In this article, we demonstrate the work of a Johns Hopkins University Applied Physics Laboratory (APL) team to automate recognition of target surface vessels from these modalities on a next-generation spaceflight processor to simulate on-orbit detection. These detections can be fused to form quality tracks that can then be used to detect dark ship anomalies via pattern-oflife analysis. Tracks formed over a continental or global scale motivate the need for further automated analysis since a significant amount of human effort would be needed to analyze thousands or tens of thousands of tracks in detail and in real time. To address this challenge, the APL team developed a suite of pattern-of-life tools that extract features from tracks and flag tracks that deviate too far from some learned definition of normality.
ISSN:0270-5214
1930-0530