See Insulator in the Dark: Insulator Defect Detection under Low-light Condition based on SliD

Insulator defect detection has always been one of the important tasks of transmission line state detection and operation and maintenance. Nowadays, insulator defect detection based on universal data sets has achieved very impressive results. However, the requirements of existing detection systems for data sets remain in the normal light, and the corresponding work has not been carried out for insulator defect detection under lowlight conditions. This paper proposes a low-light detection model named See Insulator in the Dark, which makes it possible to effectively patrol under different light conditions and improves the robustness of AI-assisted patrol.

I. Introduction

Transmission lines are widely used in the industrial field and daily life in modem society. They are mainly composed of cables, insulators and other components. However, these basic components are often subject to wear and corrosion in the outdoor environment, resulting in unexpected interruption of power supply and even serious casualties. Therefore, visual inspection is often used as a method of regular inspection and predictive maintenance to ensure the reliability of transmission lines [1]–[3]. As the most commonly used power line visual inspection method, UAV patrol inspection can not only meet the requirements of UAV route planning, but also be used in the field of visual inspection and detection [4]. However, due to bad weather, inspection time, shooting angles and other reasons, the illumination condition of the original data acquired by UAVs is irregular.