Digital Image Processing Applied to the Reduction of Subjectivity on Ultraviolet Imaging Diagnosis of Insulators

Detecting and monitoring the corona effect in electrical power substations can prevent failures and reduce maintenance costs and fines. Ultraviolet imaging stands out as a tool for this purpose; even though the interpretation of its results is not trivial. This article presents a new methodology, based on digital processing of ultraviolet videos, for the automated classification of the operational condition of polymeric insulators (500 kV class). Initially, the ultraviolet representations (facula) of all video frames are extracted by color threshold. Then, a persistence map of the registered phenomenon is constructed. From this map, digital image processing and computer vision techniques are used to extract relevant characteristics of detected corona discharges, such as, intensity, persistence and distance between the source and the polymer. With the application of the proposed methodology to a wide database, it was observed that the classification parameters were more effective than those present in the bibliography. Furthermore, the developed algorithm was able to classify the operational condition of the insulators correctly and to locate the origin of the discharges. Finally, there presentation in the form of a persistence map proposed is an effective tool to summarize the useful information of an entire video in a single image.