Moving target perception and 4D attitude calculation in flight area with vehicle-road coordination

To overcome the limitations of current remote control systems that depend heavily on manual intervention and limited image detection, particularly in managing intricate aircraft coupling and addressing vehicle drivers’ imprecise aircraft position judgments, this paper presents a novel vehicle-road collaborative activity target perception and 4D pose estimation algorithm specifically tailored for traffic management and control in flight zones, such as airport runways and aprons. By leveraging the YOLOv6 framework, fused with the innovative EfficientRep Backbone and Rep-PAN Neck, the algorithm achieves direct regression of target poses, 3D spatial coordinates, and confidence scores, all without relying on prior aircraft or skeleton features. To address the scarcity of aircraft samples from vehicle drivers’ viewpoints, 3D modeling is utilized to construct an airport model and simulate a diverse multi-view aircraft dataset, enabling transfer learning from the remote control tower’s perspective to the vehicle’s. Experimental outcomes validate the algorithm’s precision in accurately computing aircraft 4D poses in both real-world and simulated environments, thereby providing robust support for vehicle-road collaboration in flight zone management, significantly enhancing safety and efficiency in these complex environments.