An Improved YOLO11-Based Algorithm for Small Object Detection in UAV Aerial Imagery

Object detection in aerial imagery presents significant challenges due to the small size, low resolution, and indistinct features of targets. While YOLO-based algorithms are effective for detecting larger objects, they struggle with the detection of small targets. To mitigate this deficiency, an improved UAV detection method leveraging YOLO11 is presented. By incorporating dynamic convolutions in place of traditional ones, our approach improves both detection accuracy and computational efficiency. Furthermore, the integration of the PKI Module and CAA modules expands the receptive field and refines feature attention, while the SPPF-ECA module enhances channelspecific focus. A FFDPN structure is also introduced to improve multi-scale feature fusion. Results from experiments show that our approach exceeds the performance of YOLO and other cuttingedge models, achieving substantial gains in mean average precision (mAP) and small-object detection.