A comparative evaluation of different segmentation techniques to detect flaws in weldments radiographically

In this paper, the concept of application of three segmentation techniques, ie morphological edge-based segmentation, region growing segmentation and multistage watershed segmentation, has been implemented on radiographic NDT images of weldments for the detection of flaws produced during the welding process. These methodologies are compared and concluded to be effective for the detection of all nine possible types of weld flaw (slag inclusion, worm hole, porosity, incomplete penetration, under cuts, cracks, lack of fusion, weaving fault and slag line) and feature extraction. After being successfully tested on more than 80 radiographic images obtained from EURECTEST, International Scientific Association Brussels, Belgium, 24 radiographs of ship welds provided by Technic Control Co, Poland, were used, obtained from Ioannis Valavanis, Greece. A comparative table showing the results is also presented in the paper. Many researchers used ANN, fuzzy logic, SVM and ANFIS novel methods to segment such types of image. Even though the results of these methods are impressive, they are time-consuming and require complex implementation. In contrast, the present methods are simple to implement and are optimised for time. The procedure to detect all types of flaw and feature extraction is implemented by a segmentation algorithm, which can overcome computer complexity problems. The experimental results show that our proposed method gives good performance on radiographic images.