A Computer Vision-Based Standalone System for Automated Operational Data Collection at Non-Towered Airports

The accurate collection of operational data at airports is essential for ensuring the fair distribution of national funds. However, many U.S. airports lack control towers, which forces planners to rely on sound, radio, and transponder-based systems for detecting aircraft operations. While these methods are useful, they have limitations, such as low accuracy and an inability to identify specific aircraft. In our previous work, we developed a computer vision-based system capable of accurately counting and identifying aircraft. However, implementing this system requires powerful computing devices that are not typically found at non-towered airports. Additionally, cloud computing is not a viable option due to data transfer limitations and associated costs. To address these challenges, we propose an affordable solution utilizing edge computing. This paper describes the necessary software and hardware modules, including optimized machine learning methods and edge devices, for deploying the system at airports. We tested the system’s performance using two independent standalone setups, created with NVIDIA edge kits, at three non-towered airports. The first setup aimed to obtain an accurate count of operations without detailed aircraft information, while the second setup was designed to extract an operations count with comprehensive aircraft information, including aircraft type recognition and identification. The accurate retrieval of aircraft information in the edge computing system is achieved through the introduction of a tailored CNN-based recognition model and the execution of a 3-step tail number identification algorithm. The results demonstrate the practical value of the proposed system, indicating that an accurate count of operations with detailed aircraft information can be obtained at a reasonable cost.