Currently, unmanned aerial vehicles (UAVs) have been extensively utilized in ground target capture tasks, accompanied by the challenge of detecting and identifying small objects in UAV-captured images. Moreover, UAV-captured images exhibit varying scales of objects, posing additional difficulties for object detection tasks. To address these challenges, we propose an enhanced algorithm based on YOLOv8. In our algorithm, we incorporate Omni-dimensional Dynamic Convolution (ODConv) into the C2f module and apply this newly designed module to the backbone of YOLOv8. This integration enables us to leverage the advantages of ODConv, which features dynamic filters across multiple dimensions. This modified backbone reduces the model’s inference time while simultaneously enhancing prediction accuracy. Additionally, we introduce a novel Small Object Detection Structure (SODS) specifically designed for small object tasks. SODS focuses on small object regions at larger feature scales, resulting in the improvement of detection accuracy of small objects. Our proposed algorithm has been evaluated on the common unmanned aerial vehicles view datasets, and the experimental results reveal a significant improvement in both mAP and Recall values.