DC Corona Discharge Mode Identification Based on the Visible Light Images via the YOLOv8

This article proposes a new approach for identifying both positive and negative dc corona discharge modes using advanced computer vision techniques. The You Only Look Once (YOLO) version eight models for corona discharge localization and deep neural networks for corona mode classification were applied. These corona discharge modes were created using three distinct rod electrodes under dc excitations. Visual representations (corona discharge light patterns) of these modes were taken by utilizing a commercially available digital single-lens reflex (DSLR) camera. Furthermore, to enhance the diversity of our dataset, corona discharge images from various literature sources were incorporated. The YOLOv8 was first implemented to locate corona discharge sources accurately within the recorded images. Subsequently, deep convolutional neural networks (CNNs) classified the different modes of corona discharges. The main contribution of the study is to generate a pioneering framework for advanced corona discharge identification using visible images captured with a digital camera. Moreover, it employs the YOLO model, which is the first attempt at corona discharge identification through visual images. Moreover, it demonstrates the applicability of the digital camera as a sensor for detecting all modes of dc corona discharges. The results show that this approach successfully pinpoints the corona discharge location and accurately classifies them.