Active Vision-Based Alarm System for Overrun Events of Takeoff and Landing Aircraft

The safe takeoff and landing of aircraft is a very important part of aviation safety. However, the tower controller cannot actively obtain the status information of the takeoff and landing aircraft to issue warning information of the overrun events so that the aircraft can go around in a timely manner. To solve this problem, this article designs an active vision-based alarm system for overrun events of takeoff and landing aircraft. First, the system utilizes the nested densely atrous spatial pyramid pooling (NDASPP) module to capture the features of aircraft with different sizes and reconstructs the 3-D skeleton of the aircraft based on it to characterize the aircraft time-space positioning information (TSPI) with six-degree-of-freedom (6-DoF) representation, i.e., position and attitude angle. Then, the 6-DoF TSPI of takeoff and landing aircraft will be estimated through the simple convolutional network. Second, flight operational quality will be evaluated through the estimated aircraft 6-DoF TSPI to achieve a diagnosis of overrun events. Finally, the proposed system uses the visualization function to show the tower controller the takeoff and landing process of the aircraft, monitoring parameters, and the diagnosis of the overrun events so that the tower controller can issue warning information. The experimental results show that the proposed system outperforms 3DSke by 9.4%, 15.3%, 20.1%, 2.7%, and 8.6% on the metric average 3-D distance (ADD), Rete, Re, Te, and 2-D Proj, respectively, and can control angular error within 1° and linear error within 1.9 m. The runtime of this system is only 67 ms. In addition, it can effectively detect ongoing or foreseeable overrun events.