Target detection for aerial images has high application value in military and civilian fields.To solve the problems of low detection accuracy and inaccurate positioning due to factors such as small size of the targets, wide scale ranges, and background interference in UAV aerial images, a target detection algorithm for UAV aerial images based on the improved YOLOv8n is proposed.Firstly, the C2f module is improved, and the Deformable Convolutional Network(DCN) is used to replace the convolution in its Bottleneck to adapt to the deformation and scale variations of the objects in aerial images.The LSK attention mechanism is introduced into the backbone to dynamically adjust the spatial receptive field, thereby more flexibly adapting to the differences in background information requirements of different targets at the feature extraction stage.Then, the neck structure is improved, a shallow detection layer is added and the big target detection layer is removed, so that the network can more effectively capture the features of small targets to improve detection accuracy.Finally, the WIoU loss function is introduced to make the model focus more on low-quality samples and obtain higher detection accuracy. Comparative experiments and ablation experiments were conducted on the VisDrone2019 dataset.The mAP₅₀ value is increased by 5.2 percentage points compared with that of the baseline model, the parameter count is reduced by 20%, and the detection speed reaches 87 frames per second, which can meet the real-time detection requirements. Comparative experiments were conducted with mainstream algorithms, and its performance is better than that of current mainstream algorithms.A generalization experiment was conducted on the DOTA dataset, and the mAP₅₀ is increased by 1.7 percentage points, proving that the algorithm is versatile.