A novel constraint-based data fusion system for limited-angle computed tomography

Computed tomography (CT) is a non-destructive evaluation technique that reconstructs the cross section of a specimen from x-ray raysum measurements. Whereas CT reconstruction is an ill-posed inverse problem that is easily solved, limited-angle CT, where raysum data are missing for a range of angles, is more severely ill-posed and more difficult to solve. In the limited-angle case, a priori assumptions are necessary to constrain the problem. Specimens wider than the x-ray source to sensor spacing require limited-angle CT. Furthermore, if the specimen is a sandwich structure, i.e., some core material surrounded by load-bearing face sheets, then the face sheets must lie in the null space. Components in the null space do not appear in the raysum data and thus confound CT reconstruction because there is no basis for interpolation. This thesis proposes a novel constraint-based data fusion method for limited-angle CT reconstruction of sandwich structures. The method reduces the reliance of limited-angle CT on assumptions by using range and ultrasound measurements to constrain the solution. Fusion of the data sources results in a problem with a much smaller null space that no longer includes the face sheets. The reduction of the null space in a manner consistent with the specimen yields a more accurate tomographic reconstruction. Synthetic and real data experiments show marked improvement in reconstruction accuracy achieved by using the fusion system.Science, Faculty ofComputer Science, Department ofGraduat