Automated gas pipeline corrosion detection with artificial intelligence

The article presents a method to detect gas pipeline corrosion using artificial intelligence to analyse visual images with 3 steps: preprocessing of input images; segmentation and extraction of histogram features and texture features; and proposing to use the hidden Markov model trained from feature vectors capable of automatically analysing the camera images and identifying eroded areas of the gas pipeline. An initial experiment on a dataset of over 5000 published oil pipeline images shows the proposed method achieves results with over 90% accuracy.