Automatic NDT analysis of x rayed welding images

Digital image processing is commonly used in nondestructive testing (NDT) to improve the quality - i.e., legibility - of x-rayed images. In the case of weld radiography, however, the subsequent critical steps of image analysis and defect location classification are normally carried out by human operators, who are inevitably prone to subjective diagnosis errors. This work presents software designed by the authors to automatically analyze x-rayed weldings, which may be adapted to any database. In experiaental testing, the algorithms developed were successful in automatically locating and classifying defects on radiographic images.