AI-Enabled Multi-Sensor Fusion for Combat Search and Rescue

RT&L Focus Area: Artificial Intelligence & Machine Learning -- When a downed pilot goes into the water, he is on borrowed time before exposure can cause hypothermia. Recognizing this, the Air Force operates dozens of Rescue Squadrons around the globe, whose Airmen put themselves into harm’s way to save lives of fellow warfighters and public. During search and recuse (SAR) operations, flight crews collect hours of full motion video and thousands of square kilometers of imagery, despite using just a few aircraft. As minutes pass, currents displace isolated personnel from where they crashed, increasing exponentially the search box area. Flight crews and analysts are overwhelmed by the volume of work, leaving too few areas searched and too much imagery unanalyzed.   These dual problems, too few SAR aircraft and too much imagery data, can be addressed by leveraging the Large Aircraft Infrared Countermeasure (LAIRCM) or other sensors already deployed on aircraft across the Air Force, and using artificial intelligence (AI) built by CrowdAI to automate imagery analysis. As an established program of record across multiple services, LAIRCM is not only well suited programmatically for this project, but its infrared sensor is particularly adept for open ocean search. What is more, LAIRCM sensors are incorporated into across the DoD aircraft fleet, as well as in operation with our allies. The ubiquity of this sensor platform, when equipped with automation tools, could expand exponentially the SAR capabilities of the United States, with only a modest investment in AI. This computer vision solution could reduce the time and resources needed to scan imagery to find and recover isolated personnel before succumbing to exposure.   Under Phase I award, we will provide a product demonstration with 920RW, AATC, and other potential users/customers. CrowdAI has identified AFRL Rescue Vision as the likely route for integrating our solution into the Air Force enterprise. Working with the 920th Rescue Wing, Air National Guard Air Force Reserve Command Test Center (AATC), Air Force Special Operations Command, and Air Force Research Lab Rescue Vision program, CrowdAI will build custom AI models to process the LAIRCM or other imagery in real-time from on-board a rescue aircraft. CrowdAI has developed best-in-class deep neural networks (DNN) that generate industry leading machine learning (ML) models for automated imagery exploitation, or Computer Vision (CV). CrowdAI ML models solve specific use-cases, exploiting virtually any form of literal (e.g. IR, pan-MSI) or non-literal (e.g. SAR) imagery collected by ground, aerial, and satellite platforms—including full motion video and still imagery sensors. These models can evaluate greater quantities of imagery than an analyst, but in a fraction of the time—and AI never tires, reducing the probability of error.