SERPONG — Industrial Computed Tomography (CT) utilizing Gamma-ray radiation has long been a cornerstone of non-destructive testing (NDT), offering critical cross-sectional insights into thick-walled structures such as subsea pipelines, pressure vessels, and heavy industrial machinery. However, performing complete 360-degree rotational scans during real-world, on-site field inspections remains a major operational challenge. Physical obstacles, complex geometries, and tight accessibility constraints frequently prevent detectors from capturing a complete set of projection angles.
To solve this industry-wide challenge, our research team, in active collaboration with university researchers, is developing a novel Deep Learning model designed for missing-projection recovery in Gamma CT sinograms.
Overcoming Partial-Data Constraints with AI In standard Gamma CT reconstruction, missing angular projections lead to severe image artifacts, loss of spatial resolution, and inaccurate structural analysis. Our deep learning framework targets this core issue at the raw data level. By training specialized neural networks to analyze contextual features within incomplete sinograms, the model accurately predicts and interpolates the missing projection data prior to image reconstruction.
This approach enables high-fidelity, artifact-free cross-sectional imaging—even when physical field constraints limit scan coverage.
Key Technical Advantages Sinogram Inpainting: Reconstructs missing projection angles directly in the Radon domain before CT reconstruction algorithms run.
Enhanced Field Feasibility: Eliminates the need for costly mechanical modifications or dangerous scanning maneuvers around physical obstructions.
Maintained Measurement Integrity: Preserves density mapping accuracy for internal defect detection, corrosion monitoring, and wall-thickness evaluation.
Call for Academic & Research Collaboration The project is currently in its active model development and validation phase. As we advance toward experimental benchmarking and real-world field trials, we are actively seeking collaboration with academic institutions, research labs, and industry partners specializing in:
Advanced CT image reconstruction algorithms
Industrial non-destructive testing (NDT) & Gamma radiography
Applied machine learning for inverse imaging problems
Interested in collaborating?
Institutions and research teams interested in joint research, benchmarking, or technology transfer are invited to reach out to our project lead at fahr006(at)brin.go.id