An Investigation of Automating Fungus Inspection Process of Aircraft Fuel Tank via Image Processing

Fungus growth in the fuel tank can be harmful to the safety of an aircraft due to its corrosive nature and the sludge it produced. Thus, maintenance is often conducted within a definite period of time by draining the fuel tank to inspect the presence of fungus manually via limited access points. This hinders the full view of the tank resulting in the undiscovered fungus growth, which is contrary to the aim of the inspection, at the expense of resources. This study aims at automating the inspection process to detect the presence of fungus colonies in the aircraft's fuel tank based on the camera's image. It was observed that fungus colonies are often formed irregularly surrounding the bolts of the tank's inner structure. This makes it challenging to differentiate as the color of bolt's edge and fungus colonies is similar. The proposed algorithm aims at addressing the challenge through background and edge removal by Gaussian filtering, adaptive thresholding, convolution for eliminating rogue pixels and boundary calculation. It was tested against the images taken from both experimental rig and aircraft's fuel tank, where the algorithm detected the fungus colonies from the experimental rig with 100% accuracy. In contrary, there were several false detections observed in detecting the fungus grown in the aircraft's fuel tank but it is still satisfactory.