Identification And Defect Detection Of Vibration Damper On Power Transmission Lines Based On YOLOv7

To ensure the normal transmission of electricity,regular inspections of transmission lines are necessary.The wires hanging in the air for a long time are subjected to wind-induced vibrations,which can lead to metal fatigue and impact the safe operation of the power system.The installation of vibration damper aims to reduce the impact of wind on the wires.With the rise of smart grids,drone-based intelligent inspection methods are gradually being used,and the research on detecting vibration damper in the collected images is of significant value.However,the wide field of view and complex backgrounds in the images captured by drones at high altitudes increase the difficulty of detecting vibration damper.Therefore,this article integrates power line inspection with visual processing technology to study the recognition and defect detection algorithms for vibration damper on transmission lines,aiming to improve the accuracy and efficiency of power line inspections.Therefore,this article combines power inspection with visual processing technology to study the identification and defect detection of vibration damper on transmission lines.A dataset for seismic damper detection is established for algorithmic research.To improve detection speed and reduce model size,a lightweight pointwise convolutional network is introduced.Different versions of Shufflenetv1 and Shufflenetv2 are respectively used as the backbone extraction network of the single-stage detection network YOLOv7 algorithm for seismic damper recognition experiments.Compared to Shufflenetv1,the Shufflenetv2-YOLOv7 algorithm model has higher accuracy in identifying vibration damper.To address the problem of multi-scale and dense vibration damper that are difficult to recognize due to shooting angles in aerial images,a lightweight attention mechanism module CBAM is introduced,and the seismic damper recognition experiments before and after the algorithmic improvement are compared,as well as performance comparisons with commonly used algorithmic models to verify the effectiveness of the method.The improved object detection network is also used to detect rust defects on vibration damper,and the detection speed is further improved by introducing a decoupling head.Tests on the self-built seismic damper dataset reveal that the improved network can effectively detect the vibration damper and defects,with an average precision of 93.1%,which is improved by 1.5% compared to Faster R-CNN and SSD algorithmic models.In seismic damper defect detection,the accuracy of defect recognition is 92.4% after introducing the decoupling head,an improvement of 0.7%.The performance of the improved algorithm is good and can provide assistance in seismic damper inspection work.