A Dual Fine Grained Rotated Neural Network for Aerial Solar Panel Health Monitoring and Classification

Abstract

The paper suggests dual two-staged novel fine grain rotated network for aerial solar panel health classification. The neural network architecture can detect different types of uncleared solar panels of any arbitrary orientation installed in various environments. Three different types of solar panels were identified and categorized based on accumulation of dust. A synthetic dataset was generated to assess the precession of the solar panels’ detection in various aerial spatial situations. Furthermore, no datasets were available to support this research. An amalgamation of two datasets were used to draw a conclusion. We primarily focus on identifying different types of dirty PV panels, and we specifically address dust accumulation. The model proposed in this work is effective because it is extremely sensitive to the collection of dust which are hard to detect from an aerial field of view.