Detection of Rust from Images in Pipes Using Deep Learning

Abstract

This paper proposes a method for detecting rust inside pipes using deep learning. In recent years, the number of pipes that have passed their useful life has been increasing, and earthworm-type robots have been developed to perform regularly inspections of sewage pipes. The images of the sewage pipe taken by the robot are trained on a Variational Auto Encoder, which is an unsupervised learning model, to detect abnormalities by taking the difference between the input image and the output image. In addition, the trained Residual Network is used to estimate the location of anomalies.