Computer Vision for Tactical Alerting (CV-TA) Phase II

The space domain has been seeing a rise in congestion of artificial Resident Space Objects (RSOs). This is compounded by the proliferation of mega constellations like Starlink by SpaceX and increases the probability of conjunction events and collisions. Not only is this problematic for the commercial sector, but detrimental to our defensive space infrastructure and assets as well. Our adversaries understand the importance of the space domain to our national defense and could potentially target these assets to degrade our Command, Control, Communications, Computers, Intelligence, Surveillance, and Reconnaissance (C4ISR) capabilities (e.g., ambiguity of the purpose of China’s Shijian-17, -21 satellites). The current alert infrastructure generates large amounts of on-site data. It is not practical or possible to transmit every image from these facilities, as the data volume can quickly burden the available transmission bandwidth. Instead, near-real-time tactical alerts are generated because only partial images/centroided detections are typically transmitted. However, this implies that threat/conjunctions events can be missed, mischaracterized, or may go undetected, sometimes for days or weeks. In addition, manual visual inspection of every image is an ever intractable, and impossible task. The alerts that are generated – for instance in the cases of closely spaced objects (CSOs) or potential break-up events – require validation and verification through follow-up imagery. Operators must wait for a potential alert to go through the tasking, collection, processing, exploitation, and dissemination (TCPED) chain before requesting follow-up verification of the event. The AMC-9 break-up event is an example of the current space-domain awareness (SDA) architecture. But the current alert-infrastructure does not scale with the growing congestion of cislunar RSOs, associated threats, and conjunction event notifications. The USSF critically needs a real-time, automated process to detect conjunction events (CSOs, break-ups, etc.) and in real-time to alert operators for COA planning and execution. We propose a computer vision solution that utilizes non-resolved RSO imagery prior to object correlation to produce tactical information for the warfighter.