ABSTRACT
A three-step procedure is described which utilises wavelet de-noising, shape from shading (SFS) method and the Harris corner detector (HCD) for improved defect detection from radiographic images of welded objects. As the first step, the digitised radiographic images were processed using wavelet de-noising to remove the random high frequency noise originating predominately from the detection of scattered radiation. The SFS method was then applied to distinguish and discount low gradient morphological features from the higher gradient defect regions. Finally, the HCD method was applied to detect and highlight the remaining edges and corners associated with different weld defect types. Results from the application of the three-step procedure have shown unanimous agreement between operators as to improved weld defect detection compared to the unaided manual methods. Utilising artificial intelligence techniques, the procedure has the potential to be extended towards automation of defect detection with reduced operator supervision/intervention enabling increased testing throughput without compromising defect detection quality. It is further suggested that the procedure may also be developed further to distinguish between different defect types using the density and shape of the detected corners.
Disclosure statement
No potential conflict of interest was reported by the authors.