Utilizing digital image processing techniques to evaluate the condition of non-ceramic insulators
The aim of this paper is to develop an automated system to classify and assess the surface condition of silicone rubber material. Both Radon transformation and the gray-level co-occurrence matrix were examined as image processing and features extraction techniques while using the artificial neural network as a classifier. A database comprised of 358 images was collected and preprocessed representing the well-known seven hydrophobicity classes.
A recognition rate of 95.67% was achieved while using combined features from both techniques using stepwise regression as feature selection technique to form the input feature vector. The developed system overcomes the disadvantages of the current evaluation techniques by eliminating the human intervention.