Enhancing the Efficiency of Defect Image Identification in Computer Decoding of Digital Radiographic Images of Welded Joints in Hazardous Industrial Facilities

Abstract

—This article is devoted to enhancing the efficiency of defect image identification in computer decoding of radiographic images. The work addresses the task of defect image segmentation, as well as models for defect image segmentation on radiographic images in both manual and computer decoding. The distinction between algorithms for detecting and identifying groups, clusters, chains of pores, slag, and metallic inclusions in comparison to manual decoding of images is demonstrated. Algorithms for defect detection and identification for use in digital radiography systems have been developed and experimentally tested on the APC KARS system. The convergence of results between computer and manual decoding reaches 0.85.