Change Detection on Pipe Outer Surface via Reference Image Extraction from Past Video Using Feature Points

Abstract

In this study, we propose a method for detecting changes on the outer surface of pipes using inspection videos captured by a patrol inspection robot. In patrol inspections, it is important to detect changes on the outer surface of pipes by comparing them with their past normal state. Therefore, comparing the videos before and after the changes is crucial for change detection. In the proposed method, we extracted comparison image pairs using image features from videos captured by a camera on a mobile robot and performed change detection using deep learning. This method achieved a detection accuracy of 93.9% for changes in the inspection videos.