With the rapid development of unmanned aerial vehicle (UAV), there are more and more incidents of UAV losing control. The frequent occurrence of “black flights” and “indiscriminate flights” caused by improper use of UAV has seriously endangered public safety and personal privacy. Therefore, object detection technology of UAV has attracted wide attention. However, the small size of UAVs, complex airspace background and other problems still brought great challenges to this field. Based on this, we proposed an improved YOLOv8 algorithm (GF-MDH YOLO) based on gated fusion and multiple detection head for UAV detection tasks. Firstly, a high-resolution detection head is added into the detection head component to improve the detection ability of the network for small objects, meanwhile, the deepest detection head and the corresponding network layer are eliminated, which effectively reduces the parameters of the model. Secondly, the gated fusion mechanism is introduced into the neck of YOLOv8 to improve the detailed information of deep features, thus improving the detection accuracy. The experimental results on self-made data sets show that the detection accuracy of GF-MDH YOLO is as high as 89.8%, which is 6.3% higher than native YOLOv8.