An Improved UAV Image Small Target Detection Method Based On YOLOv8

Aiming at the problems of low precision, false detection, and missing detection in UAV images, an improved small target detection method based on YOLOv8 is proposed. Firstly, a spatial depth conversion convolution SPD-Conv is inserted in front of each C2F layer in the backbone to improve the algorithm’s generalization on low-resolution images and feature extraction ability for small targets, which can make the algorithm better adapt to the complex and changeable situation of small and medium-sized targets in images. Secondly, in the detection head part, the structure of the detection head is improved, and the small target detection ability and feature fusion ability of the model are improved by adding a small target detection head and deleting large target detection head, and the number of parameters of the model is controlled to reduce the calculation time of the model. The comparison experiment on the VeDAI dataset proves that the mAP of the algorithm can reach $84.7 \%$, and the prediction effect on the subjective result graph can prove that the improved algorithm has a good effect on small target detection, and can improve the algorithm’s target detection performance on UAV images, which has certain feasibility and effectiveness.