Previous Abstract Return to Session F6 Next Abstract

Session F6: PNT for Uncrewed Systems

Maximum Range of Computer Vision for Aircraft Classification
Ryan Stiffler, Scott Nykl, Air Force Institute of Technology
Location: Room 6-8
Date/Time: Tuesday, Jun. 4, 11:50 a.m.

Automated Aerial Refueling (AAR) has emerged as a forefront area of exploration and innovation within the Air Force, aiming to empower autonomous drones to independently refuel during prolonged missions. The intricate challenge of AAR unfolds into various research facets, and this paper zeroes in on a crucial aspect: aircraft classification.
In the intricate process of refueling any aircraft, fast identification of the refueling plane is paramount, guiding the refueler on the specific procedures to employ. To expedite this task within an autonomous drone and refueler scenario, leveraging a camera to transmit image data to a trained neural network proves instrumental in classifying the aircraft swiftly and accurately.
Neural networks showcase remarkable proficiency in discerning diverse objects, even at minimal image resolutions. This paper's objective is to ascertain the neural network's capability in accurately classifying a given aircraft at maximum distances or minimum resolution. As an aircraft recedes from the camera, the diminishing number of pixels in the image poses a challenge, making differentiation more intricate. The pivotal aim is to determine the earliest point at which the neural network reliably classifies the aircraft.
This point signifies the maximum range for aircraft classification, predominantly contingent on the neural network's capabilities. Nevertheless, several external factors, including camera distortion, frame obstructions, weather conditions, and darkness, exert an influence on the system's functionality. However, for the scope of this research, attention will not be diverted to these challenges, and the assumption is that the system operates under normal conditions.

Previous Abstract Return to Session F6 Next Abstract