Apron Activity Understanding based on Video Perception

Accurate understanding of apron activities is crucial for intelligent airport management and control, and one of the key tasks is to identify key nodes in the flight turnaround process so that they can be compared with the expected schedule, promptly adjust subsequent support tasks and optimize resource allocation. Currently, most airports still adopt inefficient manual recording methods, which suffer from poor real-time performance and accuracy. Moreover, existing research on flight turnaround node identification is scarce, lacking comprehensive induction and analysis. This paper systematically reviews the key apron activities and proposes a video perception-based framework for identifying key apron activities, which can integrate node feature detection results from multiple video streams for comprehensive judgment. Additionally, we designed a delay confirmation mechanism based on dynamic frequency statistics, which is applied in four types of activity analysis algorithms, and a mask filtering mechanism is introduced to collectively enhance node identification performance. Experiments show that the system can automatically identify 28 nodes, significantly improving node collection efficiency and accuracy compared to traditional methods.