Research on Two Types of Defect Detection of Transmission Tower Based on Improved YOLOv5s
With the rapid development of domestic electric power industry,the defect detection and repair of transmission towers are key technical means to ensure the safe operation of the power grid.At present,it is mainly to identify the defects of transmission tower manually,and the work burden is huge.Therefore,we propose an improved YOLOv5s object detection algorithm based on the YOLOv5s network to improve the efficiency of detection.Focal-EIoU loss function is introduced into the basic model to improve the convergence speed and accuracy of the model.The Hardswish activation function is introduced into the convolution layer to improve the expression ability of the model and the precision.The confidence threshold conf-thres of the algorithm reasoning is increased to reduce the false de-tection of model reasoning and improve the positive detection rate of the model.In addition,in the research,we tried to integrate attention mechanism to improve the ability of network feature extraction,but the effect was not ideal,so we abandoned this improvement strategy.The experimental results show that all indicators of the improved model have been improved,with a precision increase of 2.06 percent from 92.96%to 95.02%;the recall rate has increased from 87.36%to 87.38%;mAP@.5 ∶ mAP@.5 ∶.95(0.1 ∶ 0.9)increased from0.644 3 to0.648 1,an increase of 0.38 percent;the model detection speed FPS has been improved by 4.4.
YOLOv5stransmission towersdefect detectiondeep networkloss function
万方数据
维普
