3D image Reconstruction from low energy fuzzy and noisy X-ray projective data.pptVIP

* V.Vengrinovich, S.Zolotarev and Y.Reshetovski Institute of Applied Physics, Belarus National Academy of Sciences, Minsk, Belarus veng@iaph.bas-net.by IAP V.Vengrinovich, S.Zolotarev and Y.Reshetovski 17WCNDT Problem formulation Maximum x-ray tube voltage – 450 kV sufficient to radiograph steel 50-60mm This contradicts to the pressing industrial demand: TO VISUALIZE AND TO CONTROL FABRICATION OF COMPLEX INDUSTRIAL PARTS AND ENTIRE DEVICES ONSTREAM TO GET UNIFORM RECONSTRUCTION QUALITY FOR BOTH THIN AND THICK WALLS, AVOIDING RESPECTIVELY SMEARING AND SATURATION EFFECTS IAP V.Vengrinovich, S.Zolotarev and Y.Reshetovski 17WCNDT Problem formulation 1. In this elaboration we reduce the problem of image reconstruction (IR) given low dose low energy x-ray source to the problem of IR from limited, noisy and fuzzy data 2. Regular techniques for IR from limited data like: Bayesian image reconstruction using statistical image properties in the form of Gibbs priors and Kalman filtering are incapable in the case of low energy sources. L H W WD L,HD D – penetration depth IAP V.Vengrinovich, S.Zolotarev and Y.Reshetovski 17WCNDT Hull 3D image reconstruction of binary structure industrial objects represented by CAD imade: Forward Problem Solution Forward problem in hull formulation: - projection of the triangular surface element on the n-th plane - pixels which are inside the triangular - ray sum for rays within the pixel Finally the forward operator is calculated The functional measures the integrated discrepancy between ray sums of calculated and unknown surfaces, which average value is: IAP V.Vengrinovich, S.Zolotarev and Y.Reshetovski 17WCNDT Inverse problem solution is based on the minimization of the functional: where first and second terms in the right part are calculated and measured ray sums in the projections while is the average discrepancy. After this each node is shifted by the value