Automatic Target Recognition of Personnel and Vehicles from an Unmanned Aerial System Using Learning Algorithms

Black River Systems Company is pleased to submit this Phase II SBIR proposal to design, develop and demonstrate our REcognition with Aerial Learning (REAL) system. This work lays the foundation for a learning-based radio frequency (RF) unmanned aerial system (UAS) signal classifier which can address specific challenges in existing expert-driven detectors and has sufficiently low size, weight, and power (SWAP) requirements to be deployed on UAS. Our Phase I approach successfully accomplished signal classification using orders of magnitude less data and was resilient to signal structure changes where library based approaches struggle. For this Phase II effort we plan to extend and refine the machine learning models developed in Phase I; collect RF signals data representing a broad set of signal types, environments, interference sources, and target mobilities; develop novel approaches for identifying unknown signal types; deploy the REAL system on a prototype UAS platform; and perform a series of demonstrations of complete system performance.