Oil Film Detection Under Solar Irradiation and Image Processing

Abstract

To detect oil leakage from a power transformer at its initial stage to prevent environmental pollution, we aimed to detect and localize leaked oil film that had adhered to the surface of a power transformer. Since the inspection for oil leakage is generally carried out in the daytime, in this research, a method to localize the oil film under solar irradiation was developed to support the inspection. To detect oil leakages, observation of the emitted fluorescence when the oil is irradiated with ultraviolet (UV) light has been commonly adopted. However, the fluorescence of an oil film adhered to a surface is too weak to recognize under solar irradiation. Therefore, the detection of strong specular reflection light of the oil film when it was irradiated with a light containing a visible component was used to additionally detect leaked oil film. Moreover, image processing using deep learning was applied on the obtained images to automatically recognize the oil film. By using the proposed oil-localization method, the oil-adhered area on the surface was recognized with a high accuracy of 99.2%. The proposed method is expected to be used to detect oil leakages for various types of oil filled equipment.