This Optimized Detection for Space Objects effort uniquely combines utilities, techniques and processes to significantly reduce computing resources required for detection and to train high performance object recognition models. Optimized detection enables new space intelligence to be put into action sooner and reduces response lag time. The success of this effort supports Air Force missions with space object detection and tracking components; including, but not limited to: Intelligence, Surveillance, and Reconnaissance (ISR) Space Control (SC) and Space Domain Awareness (SDA) Space and Terrestrial Environmental Monitoring (EM) Command and Control (C2) Satellite Operations (SATOPS) Positioning, Navigation, and Timing (PNT) Missile Warning, Missile Defense, Kill Assessment, and Attack Assessment Space Access (SA), Space Enablers and Space Resilience. The prototype set advanced from this effort, referred to as DeLTA⢠for Space Object Detection, addresses the Department of Defense (DoD) and industry recognized challenge of costly - in terms of time and computing resources - Deep Neural Networks (DNNs) and other multilayered model training performed to achieve optimal precision and recall. Moreover, because generality across multiple sensors, convolution neural networks (CNNs) and data sets is considered as a fundamental element of the approach, the resulting prototype principles can be applied to various other markets. The near-term impact of a successful prototype effort would drastically improve object recognition system development timelines, cycles and processes in the space arena. The mid-term impact is the availability of actionable space object detection intelligence sooner. The farther-term impact is widely applicable reduction in bandwidth requirements and related computing resource stress
