Classificação da hidrofobicidade em isoladores elétricos empregando o guia STRI e processamento digital de imagem.

Abstract

The Swedish Transmission Research Institute (STRI) guide classifies the surfaces of insulators according to the contact angle and / or amount of wetted surface with water subjectively (spray method). In this work a tool is proposed for the automatic monitoring and classification of the hydrophobicity of polymeric insulators. Computational algorithms for segmentation and classification of hydrophobic images using Digital Image Processing (DIP), Probabilistic Density Function (PDF) and Artificial Neural Network (ANN) were developed. Hydrophobicity was determined using parameters of the hydrophobic images obtained by means of the spray method. The classification parameters used were: quantity of wet regions; maximum individual area; total area of wet regions; average and maximum distance between wet regions; greater form factor; and minimum and maximum eccentricity. From the obtained results a mathematical relationship between some parameters and the hydrophobicity was observed, being possible to define the hydrophobicity with a minimum set of parameters. The analysis and classification of hydrophobicity was performed using PDF and ANN. The performance was evaluated in a dataset with more than 450 images and obtained an accuracy rate of approximately 87% with ANN and of 80% by means of threshold analysis from the PDF.