Dynamic UAV Tracking Based on Adaptive Particle Filter and Its Application in Complex Airspace Surveillance

To address the challenges of UAV dynamic tracking in complex airspaces-such as tree occlusion and abrupt trajectory changes-this study proposes an adaptive particle filter algorithm (APF-ES) integrated with environmental perception. By establishing a dynamic weight evaluation mechanism, the algorithm autonomously adjusts particle counts and iteration frequencies based on background complexity, effectively resolving the issue of target loss in traditional sequential particle filter (SPF) algorithms under sudden occlusion scenarios. Experiments on a UAV inspection video dataset containing building occlusions and meteorological interference demonstrate that APF-ES reduces localization error by $42.7 \%$ compared to conventional methods while requiring only 20 particles. This provides an efficient solution for smart city low-altitude security systems.