Vehicle Target Detection from the Perspective of UAV Aerial Photography Based on Rotating Box

Nowadays, UAV aerial photography has become one of the important means for people to obtain information. The research on aerial vehicle target detection technology is of great significance to the fields of military investigation, intelligent transportation, intelligent security and so on. However, different from the ground view, the scene from the aerial view of UAV usually has the characteristics of complex background, more small targets and different target scales, which brings great challenges to the visual detection algorithm. In order to improve such problems, this paper proposes an RST-YOLOv5s algorithm. The algorithm takes YOLOv5s algorithm as the main body, designs a vehicle target detection method from the perspective of UAV aerial photography based on rotating frame regression, and improves its network structure by adding Swin Transformer. The aerial photos taken by UAV at different altitudes are made into data sets for training to realize the accurate detection of ground vehicle targets. The experimental results show that this method achieves better detection effect in the detection of ground vehicle targets by UAV.