Target Tracking Based on Multiparameter Adaptive Adjustment for Autonomous Aerial Refueling

Autonomous aerial refueling (AAR) can improve the aircraft's combat effectiveness, and drogue visual tracking is one of the key technologies for the success of AAR. Target tracking is challenging as high-speed drogue often undergo abrupt motions, leaving the vision field, and significant hollow background variation over time. To address these challenges, a tracker based on multiparameter adaptive adjustment is proposed. First, we propose a background awareness module to perceive the hollow background of the drogue, which contributes to fully utilize background information. Second, the parameter and feature adaptation module is proposed to optimize the parameters and features by combining the information of drogue motion and the changing background. Finally, the tracking result is determined by the designed adaptive fusion module that conduces to the elevation of the tracking accuracy. The three modules are effectively merged, allowing our system to track the hollow object robustly and leading to superior results on the drogue videos experiment in the sky compared with state-of-the-art trackers.