A Deep-Learning-Based Hyperspectral Detection Method of Porcelain Insulator Crack

Abstract

Porcelain insulator is an important insulating part of power system. It is easy to crack under mechanical action. If the crack is not found in time, the crack will gradually expand, and the insulating performance of the insulator will decline, and discharge accidents will occur easily. Therefore, online porcelain insulator cracks monitoring methods are needed to ensure the safe and reliable operation of power grid. However, most of the existing detection methods require offline detection and the efficiency is low. In this paper, a porcelain insulator crack detection model combining hyperspectral image and deep learning method is proposed. The results show that this method can not only meet the requirements of online non-contact detection, but also achieve high recognition accuracy.