A standardised methodology for UV camera video processing and metrological characterisation for corona discharge analysis in high-voltage equipment

• A standardised Python UV pipeline for corona discharge analysis is proposed. • Seven facula attributes, with two novel morphological descriptors, are extracted. • UV camera characterisation via a calibrated 266 nm germicidal lamp is introduced. • GUM uncertainty budget: U = 12.1 % (k = 2); frame integration: 65.9 % of variance. • Fisher ratio ranks facula area (13.2) and photon count (11.1) as top discriminators. Corona discharge, a partial dielectric breakdown phenomenon, in high-voltage (HV) equipment emits ultraviolet (UV) radiation concentrated in the 240–280 nm solar-blind window, detectable by specialized UV cameras. Despite wide adoption, the analysis of the resulting video footage remains subjective, gain- and distance-dependent, and non-comparable across cameras, and the metrological uncertainty of the derived attributes is rarely quantified. This paper proposes a standardised, metrologically grounded methodology and Python software application for automated UV camera video processing. The algorithm extracts seven quantitative facula attributes per temporal integration window, including two novel morphological descriptors (aspect ratio and extent), and produces qualitative JET colormap images encoding the spatial persistence of corona discharge. A companion metrological characterisation procedure, based on a spectroradiometrically validated 3 W germicidal lamp (peak at 266 nm, temporal stability CV < 0.5 %), is introduced. A complete GUM-compliant measurement uncertainty analysis identifies frame-integration repeatability as the dominant variance source (65.9 % of total) and reports an expanded uncertainty of U = 12.1 % (k = 2) for the facula area at the reference operating condition. Two commercial cameras were characterised over distances of 2–12 m and over the operating gain range of each camera (50–220 for Camera A; 60–180 for Camera B). The discriminative capability of the seven attributes for AC corona-type classification is quantified via the Fisher discriminant ratio (FDR), with facula area (FDR = 13.2) and photon count (FDR = 11.1) emerging as the most discriminative attributes. The methodology provides a reproducible, operator-independent foundation for machine-learning-based corona discharge classification.