Pollution Level Detection of Insulators based on Hyperspectral Imaging and Transformer Model

Pollution flashover continues to pose a major threat to the safe and reliable operation of power systems, particularly in environments where insulators accumulate contamination. Traditional detection methods have struggled to achieve non-contact, precise evaluations of insulator pollution levels. Hyperspectral imaging offers a nondestructive testing solution by combining image and spectral data, making it a promising approach for assessing external insulation. However, conventional machine learning techniques often fall short in achieving high detection accuracy. In this paper, we propose using a Transformer model to determine insulator pollution levels from HSI data. Through extensive experimentation, the proposed model outperforms traditional machine learning methods, achieving a detection accuracy of [insert specific percentage] on real-world datasets. These results demonstrate the Transformer model's ability to significantly improve the precision and reliability of pollution degree detection, paving the way for enhanced maintenance strategies for power system insulators.