Computer Vision Applied to Corrosion Evaluation on Pipe Radiographs

In this paper, we present a computer-vision approach developed for wall-thickness measurements of corroded pipes inspected by double-wall radiography. The main objective of the work is to optimize data collection and film interpretation using computerized image analysis. To date, the system is able to identify the inner and outer contours of a pipe and perform wall thickness measurements at each pixel of a radiograph. This will be presented by analysing pipes having different dimensions. Another important feature of the system concerns the algorithms' adaptability to pipe wall texture. In other words, image analysis is automatically refined around rough or corroded areas. This allows more precise thickness measurements and shape identification in critical sections as will be demonstrated with examples. Finally a brief overview of results exploitation will be given and compared with human interpretation. It will show the reliability of automated image analysis and the many possibilities offered by complete output analysis.