Improved YOLOv5s-based algorithm for foreign object intrusion detection on overhead transmission lines

As a critical component of the electric power transportation system, overhead transmission lines confront various risks, including foreign object intrusion. To tackle this issue, this paper proposes an enhanced target detection model, named KM-YOLO, aimed at effectively detecting foreign object intrusion on overhead transmission lines.The KM-YOLO model builds upon the improved YOLOv5s algorithm as its foundation. It introduces the fusion of the GC module and the C3 module to create the C3GC attention mechanism, which is integrated into the backbone network. Additionally, the paper introduces the WZ Dynamic DeCoupled Head as a decoupled detection head and adopts the SIoU loss function for the model.Experimental evaluation conducted on the home-made dataset KC-dataset reveals that the average accuracy of the KM-YOLO model improves from 0.845 to 0.894. This improvement demonstrates the significant enhancement of the model's capability in detecting foreign object intrusion on overhead transmission lines.The research methodology presented in this paper offers a practical solution to the safety detection problem associated with overhead transmission lines and has yielded substantial progress in experimental results. These findings possess considerable reference value and can offer insights and inspiration for research and practical applications in related fields.