Enhancing Airside Monitoring: A Multi-Camera View Approach for Aircraft Position Estimation for Digital Control Towers

A digital tower offers a cost-efficient substitute for traditional air traffic control towers and is anticipated to deliver video-based surveillance, which is especially beneficial for smaller airports. To fully unlock the potential of digital tower, sophisticated computer vision algorithms are pivotal for efficient surveillance. While current research predominantly concentrates on tracking aircraft movements on the airport surface, an equally crucial aspect lies in real-time monitoring of aircraft as they are on finals. This capability plays a central role in enhancing both airport and runway operations. In this context, this study introduces a deep learning approach for precise estimation of the position of incoming aircraft, covering distances of up to 10 nautical miles. This approach surpasses the constraints of monoscopic techniques by leveraging multi-view video feeds obtained from digital towers. It combines Yolov7, an advanced real-time object detection model, with auxiliary regression and autocalibration, allowing real-time tracking and feature extraction from different camera viewpoints. Furthermore, we propose an ensemble approach utilizing an Long Short-Term Memory model to combine input vectors, resulting in precise location estimation. Importantly, this method is designed to seamlessly adapt to different camera setups within digital towers. Its performance is evaluated using simulated video data from Singapore Changi Airport, showcasing stability in various scenarios with minimal predictive errors (Mean Absolute Percentage Error = 0.2%) over a 10 nautical mile range in clear weather conditions. These capabilities, when implemented in a digital tower setting, have the potential to significantly improve the controller’s capacity to coordinate runway sequencing and final approach spacing, ultimately enhancing airport efficiency and safety remarkably.