FU-YOLOv8n Algorithm for the Detection of Multiple Attachments in Transmission Lines

Insulator target detection has a wide range of application scenarios in the power system, such as automatic detection of insulator breakage, missing, dirty or flashover traces by drones carrying visible light or infrared cameras. Real-time detection of porcelain insulator cracks, composite insulator aging, and prevention of equipment failure. Aiming at the current problems of misdetection and omission in the target detection of multiple attachments in transmission lines, this paper proposes a FU-YOLOv8n algorithm for the detection of multiple attachments in transmission lines. The improved algorithm adds a small-target detection head on the basis of using WIOU as the loss function, so that the network model is more focused on small-target attachments; secondly, it uses Deformable Convolution (Deformable ConvNets v2,DCNv2) to replace the ordinary convolution in the C2f module, so that the model can better capture and sense the features of the detected targets; and finally it uses the Focal Modulation (FM) to design a Focal-SPPF module instead of the SPPF module of the original model. On the transmission line dataset, the mean Average Precision (mAP) of the algorithm is improved by 3.7% compared to the original algorithm, and the precision and recall are increased to 91.5% and 90.3%, respectively.