基于改进YOLOv8s的航拍小目标检测算法OA

Aiming at the problems such as low detection accuracy,false detection,missed detection,and large number of model parameters in the detection of small targets in unmanned aerial vehicle(UAV)aerial photography,an improved YOLOv8s aerial photography small target detection algorithm was proposed.The RepViTBlock lightweight modulewas introducedto improve the C2f module in the backbone network and the neck network,and the EMA attention mechanism was introduced to further improve the C2f module in the backbone network,enhancing the feature extraction ability and reducing the number of parameters of the model.The tripartite scale sequence coding fusion module TSEF was used to reconstruct the neck network,and the small target detection layer P2 was fused and constructed,improving the detection accuracy while reducing the number of parameters.The loss function was improved by using Inner-CIoU to enhance the performance and detection accuracy of the border regression of the model.The experimental results showed that on the VisDrone 2019 aerial photography dataset,the precision P,recall R,and average detection accuracy mAP50 of the improved algorithm were 54.2％,42.3％,and 43.7％respectively.Compared with YOLOv8s,they increased by 5.7％,8.5％,and 11.5％respectively,and the number of parameters decreased by 38.7％.The improved algorithm was applicable to unmanned aerial vehicle(UAV)target detection tasks.