The ubiquity and significance of UAV image detection are increasingly evident. As this domain evolves, it has encountered challenges, especially as drones operate at higher altitudes to capture wider views, making the detection of small objects more prevalent, and the inherent resolution limitations of the equipment complicate the identification of these targets. Traditional approaches often struggle to balance the enhancement of small object detection accuracy with increases in computational costs. To address these issues, this paper introduces a novel UAV aerial image small object detection model based on the YOLOv8n framework, termed ORU-YOLO. This model is designed to optimize accuracy while reducing the number of parameters and the cost of floating-point operations. Initially, within the backbone network’s C2f module, the Bottleneck module is replaced with a Context Guided module. This alteration not only lightens the model but also enhances its capability to jointly learn local features and their contextual attributes. Subsequently, the neck structure incorporates the concept of BiFPN, with modifications to the fusion process and methods, significantly improving detection precision while reducing parameter count. Finally, a lightweight detection head is proposed to further decrease the model’s parameter requirements. Systematic evaluation of this model was conducted using the widely-used VisDrone2019 dataset. Experimental results demonstrate that the enhanced ORU-YOLO model outperforms the original YOLOv8n baseline by achieving a 3.1% increase in mean Average Precision (mAP) at a 50% Intersection Over Union (IOU) threshold, with only 38% of the original parameter count and 85% of the Giga Floating-point Operations Per Second (GFLOPs).