Automated Detection, Identification and Tracking of Unmanned Aerial Systems

To address the Army need to automatically detect and track small unmanned aerial systems (UAS) from ground vehicles equipped with distributed aperture systems (DAS) of infrared (IR) cameras, Physical Optics Corporation (POC) proposes to develop new software for Automated Detection, Identification and Tracking of Unmanned Aerial Systems (ADITUAS), based on an innovative combination of machine vision and machine learning algorithms. These algorithms will enable real-time extraction of data relevant to counter-UAS (CUAS) systems from raw DAS output. The algorithms will work on moving DAS platforms, being able to filter out platform movement before attempting UAS detection. Specifically, innovations in image registration, in using convolutional neural networks (CNN) to detect moving objects in cluttered environments, and in structure-from-motion (SFM) ranging will enable the software to detect, identify, and track UAS using DAS video output, and optionally translate UAS positions to 3D world coordinate systems. As a result, the software will provide actionable input to CUAS systems, which directly addresses the Army requirements. In Phase I, POC will demonstrate the feasibility of the ADITUAS approach by developing a Technology Readiness Level (TRL)-3 software implementation. In Phase II, POC will build a TRL 6 ADITUAS system based on COTS GPU hardware.