MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641112212 A) filed by Gowtham S; Sam Johnston C; and Sam V George on September 18, 2026, for System And Method For Automated Defect Detection And Boundary Mapping In X-Ray Weld Images.

Inventor includes Sam V George.

The application for the patent was published on September 25, 2026, under issue no. 39/2026.

Abstract: This invention describes a computer-based deep learning system built to automatically detect and segment weld defects in industrial X-ray images. The system takes in one or more X-ray weld images and runs them through a two-stage pipeline: first a primary classification network, then an advanced, modified segmentation network. The primary classification network handles the first step, sorting each image into defective or defect-free that way, defect-free images are filtered out early and never waste time in the segmentation stage, which keeps the whole system running efficiently. Any image that comes back defective moves on to the modified segmentation network, which generates precise, pixel-level segmentation masks showing exactly where the, defects are and how far they extend. This segmentation stage isn't just a standard segmentation network — it's been enhanced with 3 modules and finally it achieves 88% IOU. Working .together, these additions let the system detect and segment weld defects of all kinds of sizes, shapes, orientations, and structural complexity, while still keeping fine boundary details intact. The end result delivers accurate, efficient, fully automated weld quality inspection — well-suited to industrial non-destructive testing (NDT) work. Because it's built as an adaptable framework rather than a single-purpose tool, it can be used across a wide range of radiographic imaging setups and industries, including aerospace, automotive, shipbuilding, pipeline inspection, pressure vessel manufacturing, railway infrastructure, structural steel fabrication, power generation, and any other setting that needs reliable, automated X-ray weld defect detection and segmentation.

Disclaimer: Curated by HT Syndication.