Unmanned aerial vehicle(UAV)aerial target tracking is one of the core technologies in UAV applications.To address the problem that target appearance variations caused by occlusion

scale changes

and other factors are difficult to effectively represent using a static template

an adaptive template updating-based UAV aerial target tracking algorithm is proposed.First

a template update network integrating a Transformer basic unit and an adaptive gating mechanism is designed to dynamically update the template

so as to fully exploit temporal contextual information and enhance the template’s ability to represent target appearance variations.Second

a high-confidence criterion is introduced to filter out low-confidence sample frames

which effectively suppresses the accumulation of noisy information during template updating and reduces the risk of template contamination without introducing a large number of hyperparameters.Finally

experiments on the UAVDT and UAVTrack112 aerial datasets demonstrate the superior performance of the proposed algorithm.Meanwhile

practical deployment tests show that the algorithm achieves 40fps on the NVIDIA Jetson TX2

thereby enabling robust and real-time UAV aerial target tracking.