FCSA-YOLO: A UAV Object Detection Algorithm Integrating Frequency Dynamic Convolution and Spatial-Channel Synergistic Attention

In this paper, an algorithm named FCSA-YOLO is proposed to address the limited feature extraction capability in UAV object detection. First, to preserve high-frequency details in feature maps, conventional convolutions are replaced with frequency dynamic convolutions. Second, to enhance the network’s focus on critical regions within the feature maps, the C3K2 module is improved using a spatial–channel synergistic attention mechanism, thereby reducing the influence of redundant features on detection performance. Finally, experimental results on the VisDrone2019 dataset demonstrate significant improvements in precision and recall, as well as reductions in false and missed detections, confirming the effectiveness and practical value of the proposed approach for UAV object detection.