Insulator Overheating Infrared Image Detection Method Based on Improved YOLOv8

Abstract

Insulators are an important part of transmission lines, insulator deterioration, failure and other problems are one of the main causes of transmission line failure, and insulator deterioration, failure is often accompanied by insulators appear abnormal temperature rise, so there is a need to carry out a rapid assessment of the insulator heating condition, and can be monitored in real time at the edge end of the algorithmic approach. In this paper, to address the above problems, the improved YOLOv8 algorithm is used to improve the YOLOv8 algorithm for the insulator infrared image of the overheating region of the target recognition approach. Firstly, the YOLOv8 target detection algorithm is improved by introducing the CBAM attention mechanism, i.e., by introducing the channel and spatial attention module in the CNN (Convolutional Neural Network) to improve the perceptual ability of the overall model and improve the performance without increasing the network complexity. Then the deformable convolution DCNv3 is introduced to generate learnable offsets to improve the traditional fixed convolution operation, which realizes nonlinear sampling of the input feature maps during the convolution process and improves the robustness and accuracy of the algorithm. Finally, the WioU loss function is introduced to optimize the network, which is experimentally verified to achieve an inference accuracy of 87.5% mAP at the edge end of the device to meet the identification of abnormal temperature rise of insulators at the edge end, and the comparative test proves the effectiveness and superiority of the algorithm proposed in this paper through ablation.