Small targets in UAV aerial images occupy few pixels and exhibit sparse features, limiting the accuracy of general detection methods. This paper proposes UAST-YOLOv8, a small target detection network based on YOLOv8. To address the insufficient learning of small targets by conventional IoU loss functions, a Dynamic Adaptive Focusing IoU loss function (DAFIoU) is designed. DAFIoU employs adaptive scale-aware weighting and a dynamic focusing mecha-nism to improve bounding box regression accuracy for small targets. To reduce the high pa-rameter count of the detection head, a Lightweight Shared Convolutional Decoupled Detec-tion Head (LSCDD) is developed, which adopts parameter sharing strategies to lower model complexity. To enhance the network's focus on key target regions, a Multi-Dimensional Joint Attention module (MDJA) is proposed. MDJA integrates positional encoding with inter-head information interaction to achieve joint feature extraction across spatial and channel dimen-sions. Experiments on the VisDrone2019 and RSOD datasets show that UAST-YOLOv8 achieves mAP@50 of 40.7% and 94.4%, surpassing the baseline YOLOv8n by 7.8 and 8.7 percentage points, respectively, with only 2.3M parameters. These results validate the effec-tiveness of the proposed method for small target detection in UAV aerial imagery.