基于深度学习的输电线路外力破坏检测方法研究

Research on External Force Damage Detection Method of Transmission Line based on Deep Learning

针对输电线路附近可能出现的大型违章车辆施工造成外力破坏的情况,为保证输电线路运行的安全和稳定,提出了改进的YOLOv5目标检测算法。在原有YOLOv5算法的基础上,将其使用的Bounding box损失函数GIOU_Loss改为CIOU_Loss,使其具有更快更好的收敛效果;同时将其使用的经典NMS改为DIOU_NMS,使其对一些遮挡重叠的目标有更好的识别效果。实验结果显示,改进后的YOLOv5算法模型可以有效地监控输电线路附近的外力破坏情况。

Operation of large illegal vehicles are likely to cause external damage on nearby transmission line.In order to ensure the safe and stable operation of transmission line,this paper proposes an improved YOLOv5 target detection algorithm.Based on the original YOLOv5 algorithm,the Bounding box loss function GIOU_Loss used by it is changed to CIOU_Loss,so that it has a faster and better convergence effect.At the same time,the classic NMS used by the original one is changed to DIOU_NMS,so that it has a better detection effect for some occluded and overlapped targets.Experimental results show that the improved YOLOv5 algorithm model can effectively monitor the damage caused by external forces near transmission line.