Research on Missing Bolt Sub-Detection Based on Real-ESRGAN Super-Resolution Reconstruction

The bolt structure is one of the important accessories in the transmission line. Bolts and their missing pins account for a very small proportion of the image and the image definition is not high, so the detection of missing bolts and pins is extremely difficult. In this study, the YOLOV5 network and Real-ESRGAN super-resolution reconstruction network are integrated to adapt to the target identification job of missing bolts and their pins in intricate transmission line circumstances. The Real-ESRGAN super-resolution reconstruction network is used in this method to perform super-resolution reconstruction on photos that contain bolt targets to improve clarity. The YOLOv5 network is then used to detect the reconstructed photos, greatly improving the detection of missing bolt pins precision. The experimental findings demonstrate that the Real-Esrgan super-resolution reconstruction network in combination with the YOLOV5 network can more precisely recognize targets that were challenging to detect missing bolt pins, thereby resolving the issue of poor detection rate of missing bolt pins.

I. Introduction

In the transmission line, the bolt structure plays a crucial role in maintaining the connection and stability between various hardware and fitting components and fittings. They are the transmission line's most numerous and widely dispersed component[1]. Bolt parts have been in different outdoor environments for a long time and bear the tension of connecting hardware, which often causes missing pins. And the lack of pins will lead to the unstable structure of the hardware connected by the bolts, which will cause great hidden dangers to the normal operation of the power system. To assess and remove these possible safety issues, transmission lines are being routinely inspected by UAVs. UAV inspection not only has a high level of safety and ease of operation, but it can also analyze and gather inspection data more quickly, increasing job productivity.