UAV object tracking algorithm based on spatial saliency-aware correlation filter

Recently, correlation filter-based tracking methods have been widely adopted in UAV target tracking due to their outstanding performance and excellent tracking efficiency. However, existing correlation filter-based tracking methods still face issues such as redundant visual features with weak discriminative ability, inadequate spatio-temporal information mining, and filter degradation. In order to overcome these challenges, this paper proposes a spatial saliency-aware strategy that reduces redundant information in spatial and channel dimensions, thus improving the discriminative ability between the target and background. Also, this paper proposes a position estimation mechanism under spatio-temporal joint constraints to fully mine spatio-temporal information and enhance the robustness of the model in complex scenarios. Furthermore, this paper establishes a positive expert group using historical positive samples to assess the reliability of candidate samples, thereby effectively mitigating the filter degradation issue. Ultimately, the effectiveness of the proposed method is demonstrated through the evaluation of multiple public datasets. The experimental results reveal that this method outperforms others in tracking performance under various challenging conditions.