Autonomous Sensing of Defense Tactical Targets by LEO Imaging Systems

On this effort, Toyon proposes leveraging, extending, and developing algorithms for autonomously detecting and tracking moving aerial targets observed in space-based EO/IR data collected by imaging sensors in low earth orbit (LEO). The proposed algorithm development will enable the derivation of moving target information (MTI) and multi-track analytics based on machine learning (ML) models. Vision-based geometry will be applied to mitigate parallax and potential occlusion/un-occlusion effects caused by satellite platform motion, while ML-based approaches will be leveraged and assessed for accentuating clutter suppression and performing target/clutter classification to derive MTI data. Track-before-detect and multi-track fusion applications developed at Toyon will be leveraged for extending tracks and assessing feasibility of fusing tracks for multi-sensor information to extend target track duration. The AFIT Sensor and Scene Emulation Tool (ASSET) will be utilized for simulating sufficient data sets for model training.