To address the challenge of low detection accuracy for small objects in drone aerial images due to dense targets and complex backgrounds, we proposed MF-YOLO. First, the multi-path feature fusion capability is enhanced to integrate features from different layers, preserving shallow details and improving small object detection accuracy. Second, the EMA attention mechanism is adopted to improve the recognition rate of target regions and the accuracy, effectively distinguishing targets from background regions. Then, a Dense Attention Layer (DAL) is introduced to enhance the algorithm′s feature extraction capability in dense regions by focusing on these areas and suppressing irrelevant features. Next, a Squeeze-and-Excitation detection head is designed, incorporating the SE attention mechanism to suppress redundant features and further improve small object detection accuracy. Finally, a video dataset is constructed, and a target detection system is designed to visualize the algorithm′s detection performance. Experimental validation on the VisDrone2019 dataset shows that MF-YOLO achieves a mAP0.5 of 30.3%, a 3.4% improvement compared to the YOLOv8n baseline algorithm. The results demonstrate that the algorithm significantly improves object detection performance in drone images and has broad application prospects.