Spatiotemporal EO/IR TSPI Sensor

PhysicsAI proposes to develop a computer vision system using customized spatiotemporal deep neural networks to compute time, space, and position information (TSPI) for Within Visual Range (WVR) air combat engagements from a commercial off-the-shelf (COTS) spherical EO/IR video camera sensor. For this project, we plan to leverage proprietary spatiotemporal deep learning algorithms developed for US Army Night Vision and Electronic Sensors Directorate (NVESD) under SBIR topic A18-041 “On-the-Move Spatio-Temporal Processing and Exploitation for Full Motion EO/IR Sensor”. Spatiotemporal perception will be utilized to exploit patterns of features over time such as changes in aircraft trajectories, apparent sizes, and pose cues for greater accuracy. We plan to develop, train, and test spatiotemporal TSPI algorithms and a hardware prototype using 3D simulations, sub-scale UAV flights, and full-scale aircraft flights.