Insulator Iron Cap Corrosion Detection Based on Deep Learning

Kaicheng Guo, Yaxin Yang

Abstract: Insulators are mainly used for electrical insulation and mechanical support in power transmission systems. Because insulators are exposed to the environment and affected by leakage current, pollution and moisture, the iron cap of insulators will corrode, leading to pollution flashover accidents. Therefore, this paper proposes a deep learning based method for detecting iron cap corrosion of insulators in transmission and distribution lines. In this paper, Tensor Flow framework and Faster R-CNN algorithm are used to complete the identification task of four kinds of iron cap corrosion states of glass insulators and porcelain insulators. In this paper, the data set is augmented, the generated number and aspect ratio of anchors in RPN network are improved, and three feature extraction networks are used for comparison. The test results show that the model can effectively detect the corrosion of insulator iron caps in the line, with mAP of 83.50% and an average detection time of 0.186s per image.

Keywords: Insulator, Insulator corrosion, Faster R-CNN, Deep learning