Intelligent identification of typical defects in DR images of pipeline girth welds

X-ray digital radiography (DR) technology is widely used in nondestructive testing of girth welds in long distance pipelines. In order to improve the evaluation efficiency and avoid the influence of subjective factors of manual evaluation on the evaluation results, this paper proposed an image preprocessing method suitable for defect identification according to the characteristics of DR inspection images of pipeline girth welds. YOLO algorithm was adopted to realize the efficient and accurate identification of circular defects and unfused defects in images acquired by the same detector. The model recognition ability was verified by using engineering field images. The results showed that the method could improve the work efficiency and accuracy of the drawing evaluators, and promoted the application of DR defect intelligent evaluation technology in the field of nondestructive testing of girth welds of longdistance pipelines.