Abstract: To address the challenges of low pixel occupancy of small objects, complex backgrounds, and limited onboard resources in UAV aerial images, this paper proposes an improved small object detection model based on YOLOv8, termed YOLOv8s-SPDX. Firstly, a lightweight multi-scale neck design is introduced to the baseline network. A dedicated small object detection layer is incorporated to enhance the model's capability to detect small objects, while redundant large object detection layers are removed, effectively reducing the complexity of the network architecture. Secondly, Space-to-Depth (SPD) modules are integrated into both the backbone and neck networks to eliminate the loss of fine-grained feature information of small objects caused by strided convolutions. Finally, a C2f-MLCA module is constructed by fusing the C2f module with a lightweight Mixed Local Channel Attention (MLCA) mechanism, enabling the model to effectively combine local and global features as well as channel and spatial information, thereby further improving feature representation capability under complex backgrounds. Experimental results demonstrate that on the VisDrone-DET2019 dataset, the proposed model improves mAP@0.5 from 38.6% to 53.8% on the validation set and from 31.8% to 42.7% on the test set. On the TinyPerson and UAVDT-DET datasets, mAP@0.5 is improved by 14.5% and 2.2%, respectively. Furthermore, the number of model parameters is reduced by 20.7% compared to the baseline model.