Elaboration of novel image processing algorithm for arcing discharges recognition on HV polluted insulator model

Insulator flashover under pollution is one of the most important problems for power transmission. Occurrence of flashover is preceded by discharges propagation. This paper is dedicated to monitor discharges activity through arcing discharges pattern recognition using a combination of efficient image processing and classification algorithms. Images are extracted from recorded videos of flashover process over a plane model insulator under various contamination levels. Then, an algorithm is proposed and tested over a large image database. This algorithm processes in four stages. First, Otsu image segmentation algorithm is initially applied on images. Next, morphological filtering by combining erosion and dilation operations is computed to eliminate unwanted noises such as light reflections on the insulator model. Afterwards, connected components on filtered image are labelled enabling the calculations of four important morphological indicators consisting in the number of the connected labeled components (N_l) and the number of pixels, the length and the width of the largest connected component region (N_p, L and W respectively). These indicators characterize different properties of discharges activity and are used as an input of three well know classification algorithms (Knn, Naïve Bayes, Support Vector Machines) to distinguish between the presence or not of arcing discharges on the insulator surface. This paper introduces image processing as an efficient and fast tool for discharges activity analysis and insulator flashover monitoring. The proposed methodology dispenses the heavy instrumentations and tedious processing of conventional laboratory tests.