Research on Remote Sensing Small Target Detection Method Based on Lightweight YOLOv8

With the evolution of remote sensing technology, accurately detecting small targets in remote sensing photos has become a critical challenge in domains such as economics and defense. This study addresses the difficulty of finding small targets in remote sensing photos by offering a lightweight YOLOv8-based detection algorithm. By integrating GhostNet as the backbone network of YOLOv8, the model has been lightweighted, enhancing its operational efficiency on resource-constrained devices. Experiments with the NWPU VHR-10 dataset demonstrate that YOLOv8s-ghost has a 9% decrease in mAP but a 37.5% loss in GFLOPS when compared to YOLOv8s. This indicates that the lightweight YOLOv8s ghost maintains detection accuracy while reducing computational costs, achieving the goal of model lightweighting. This research provides an efficient solution for small target detection in remote sensing photo, with significant practical value and broad application prospects.