AI/ML-Based Development for Moving Target Engagement from LEO Imaging Systems

One of the challenges to persistent surveillance, reconnaissance, and tracking (SRT) from space is generating highly accurate moving target indications (MTI) in order to convey actionable information across the integrated battlespace. To address this challenge, advanced technologies are needed that outperform the function and operations of traditional space-based sensing systems and ground-based processing systems. One such technology, which is a pressing need for DAF, includes moving target engagement (MTE) capabilities specifically designed for adversarial maritime vessels. Toyon proposes to develop a software application that leverages advanced automatic target recognition (ATR) and target re-identification (Re-ID) machine learning (ML) models, along with mission-specific data association algorithms, to detect and track maritime vessels over long durations when data collection times between sensor revisits are on the order of several hours at best. The proposed web-based software will allow a user to choose a region of interest (ROI) in which to perform vessel detection. This will run the ATR model, overlaying oriented bounding boxes on detected vessels. A vessel may then be selected for tracking throughout a given time period. In this case, the software will overlay the track of the vessel of interest annotated with time stamps. All underlying data association algorithms will be developed to enable accurate vessel-to-database matching.