Novel image registration of a dual-camera-based corona camera for accurate insulation deterioration location identification

This paper proposes an algorithm for performing image registration in a corona camera including an ultraviolet camera and a visible light camera to detect the exact location of insulation deterioration, and evaluates the image registration error while applying the proposed algorithm to an actual corona camera. The proposed image registration algorithm calculates a linear transformation relation (Affine transformation) for image registration between a first image (ultraviolet image) and a second image (visible light image) that are simultaneously captured of a subject including multiple markers arranged radially, and aligns the images using this. The proposed algorithm achieves an extremely low image registration error of less than 0.25 mrad, approximately one-quarter of the 1 mrad specification for foreign corona cameras, enabling accurate identification of corona discharge locations.<br/>To evaluate real-time applicability, corona discharges on actual high-speed railway insulators were captured using a corona camera using the proposed algorithm. The results demonstrate that the proposed algorithm enables real-time image registration and accurate real-time identification of the location of corona discharges on the insulator (consistent with previous studies, the location of the insulator's power-side joint).