Deep4Air: A Novel Deep Learning Framework for Airport Airside Surveillance
Thai Van Phat, Sameer Alam, Nimrod Lilith, P. Tran, Nguyen Thanh Binh
Abstract
Airport runway and taxiway (airside) areas are highly complex environments, where safe-separation procedures must be maintained by Air Traffic Controllers (ATCs) under varying visibility and traffic conditions. In this paper, we propose a novel computer-vision based framework, namely Deep4Air, for automated visual monitoring of airside operations, providing real-time data including aircraft location, speed, and distance analytics. This framework includes an adaptive deep neural network that exploits a depth-wise convolutional operator for efficient detection and tracking of aircraft. The experimental results show an average precision of detection and tracking of up to 98.2% on simulated data with validation on surveillance videos from the digital tower at George Bush Intercontinental Airport. The results also demonstrate that Deep4Air can locate aircraft positions relative to the airport runway and taxiway infrastructure with high accuracy. Furthermore, aircraft speed and separation distance are monitored in real-time, providing enhanced safety management.