A Simple Model for Dirt Deposition Classification in Insulators Based on Visible Spectrum Images

In electrical plants, insulators suffer natural dirt deposition, which may eventually lead to electrical discharges, which can be avoided by the (costly) washing of insulators. Visual monitoring is a pragmatic approach to deciding when insulator cleaning is necessary. This study proposes a unique parameter based on the grayscale cumulative distribution for the classification of dirt deposition in four levels. This simple phenomenological model achieved similar results when compared to a more complex feature extraction approach for the same dataset of images. Furthermore, a simple affine formula was adjusted for a coarse Non-Soluble Surface Deposits estimation.