Ice Classification Technology for Transmission Lines Based on Visual Images

In view of the icing problem that may occur on transmission lines in cold weather, traditional manual detection methods are time-consuming and difficult to accurately identify the type of icing. To this end, this paper introduces a transmission line ice classification technology based on visual images. First, an image dataset covering five types of ice coverage was constructed to solve the problem of ice coverage type recognition on transmission lines. Then, combined with the operating environment of the anti-icing and de-icing device, a corresponding control strategy is formulated, and on this basis, an improved UNet network model is introduced. Through ice image feature analysis and Generative Adversarial Network (GAN) optimization model, the classification performance is improved. In the above data conclusions,the proposed method has a significant improvement in classification accuracy, with an average classification accuracy of 92 %. In the image segmentation task, the average value of the intersection over union (IoU) is 0.844 and the Dice coefficient is 0.89, showing high robustness and segmentation accuracy. In addition, the average absolute error (AE) of thickness calculation is 0.3 mm and the relative error (RE) is 5 %, which further verifies the effectiveness and reliability of the method in practical applications. The comprehensive experimental results show that the transmission line ice classification technology based on visual images proposed in this paper can significantly improve the automation and accuracy of ice detection, and provide effective technical support for the safe operation of transmission lines.