Drone aerial small target images suffer from defects such as overly dense detection targets, excessively small sizes, and difficulty in extracting feature information. These defects lead to low detection accuracy of existing target detection algorithms on aerial small target images. To address these issues, a drone aerial small target detection algorithm, MGF-YOLOv8, based on YOLOv8s and utilizing the Slice-Assisted Hyper-Inference (SAHI) method is proposed. Firstly, the SAHI slicing method is employed to slice remote sensing small target images, effectively mitigating the defects of overly dense detection targets and excessively small sizes. Secondly, a generalized high-efficiency layer aggregation network (GELAN) is incorporated into the Backbone part of YOLOv8 to replace the C2f module in the backbone network. This simplifies the backbone network structure, enhances feature extraction capabilities, and constructs a lightweight model. Then, a multi-scale structure (MS-block) is adopted in the neck network for feature fusion, reducing parameters while optimizing the model's performance in recognizing targets of different scales, complex backgrounds, and small targets. Finally, an additional P2 detection head is introduced to enhance the resolution of the feature map, thereby better locating small targets. Additionally, the ASFF mechanism is adopted to improve the original detection head, forming a new detection head, FASSF, to filter out information conflicts during multi-scale feature fusion and optimize the detection process, significantly enhancing small target detection capability. Experimental results on the VisDrone2019 dataset show that the mAP@0.5 and mAP@0.5:0.95 of MGF-YOLOv8 reach 56.9% and 36.3%, respectively, representing an improvement of 18.4% and 13.3% compared to the YOLOv8 algorithm. The parameter count is 10.26×10^6, a reduction of 7.82% compared to the original algorithm. The accuracy of this algorithm surpasses other similar algorithms and meets monitoring requirements, making it effectively applicable to target detection tasks on drone aerial platforms.