Corrosion Detection on Ship Hull Using ROV Based on Convolutional Neural Network

Abstract

The Remotely Operated Underwater Vehicle (ROV) has several inspection functions. One of them is the inspection function for hull damage. The damage that often occurs in the hull is corrosion. The corrosion can cause a decrease in the strength of the hull plate, reduce the speed of the ship, and decrease the quality of the safety level of ships and passengers. This study aims to classify the level of corrosion intensity on ship hulls by implementing a Convolutional Neural Network (CNN). Identification is carried out on images taken by underwater cameras via a Remotely Operated Vehicle (ROV). The intensity of the area affected by corrosion is identified so that the level of corrosion intensity can be classified and it can be considered that the ship needs maintenance to prevent even greater losses due to corrosion. The dataset used is 240 image data divided into 3 classification categories: low, medium, and high corrosion intensity. The accuracy of the real-time testing of the CNN method on the dataset plate when conditions outside the water reached 91.1% and on the dataset plate when conditions underwater reached 86.6%.