Composite insulators are used in a large number of high-voltage transmission systems in China, but because of environmental factors and other effects, the surface of composite insulators will gradually undergo insulation aging. The water repellent property of composite insulator surface is one of the indicators to reflect its insulation condition, and various methods to evaluate the hydrophobicity of insulator surface have been proposed by various scholars. The traditional method of judging the hydrophobicity level of composite insulator surface by manual work is subjective and not very efficient and accurate, while the digital image processing-based method of evaluating the hydrophobicity of composite insulator reduces the workload of the evaluation process by using computer algorithms, but the proposed characteristic parameters are still subjective. Therefore, this paper proposes a composite insulator hydrophobicity classification model based on Faster R-CNN target detection algorithm, which can automatically complete the evaluation of composite insulator water repellency, and the accuracy of this model is close to 99% with the allowed error of ±1.

composite insulators,Hydrophobicity,Faster R-CNN,BP neural network,Water spray classi-fication method