Higher accuracy with fewer projections? Automated scan angle selection for dimensional Computed Tomography based on a simple data completeness measure for the part surface

Selecting scan angles such that surface segments are aligned with straight X-ray paths (i.e., rays are tangential to the surface and therefore perpendicular to the local surface normal) is known to produce sharper transitions of those surface segments in the reconstructed volume. This enhances dimensional accuracy in sparse-view computed tomography (CT). However, existing approaches offer no direct means to exploit this criterion for automatic scan-angle optimization. We propose a method that uses a virtual representation of the CT setup, including an STL surface model of the inspected part, to automatically identify taskspecific scan angles. Using elementary vector calculus, the algorithm determines projection directions that generate tangential X-rays for targeted surface segments. To support different levels of geometric complexity, we introduce two variants of the angle-selection procedure. The methods were experimentally validated on two objects with distinct absorption and geometric characteristics. For a steel gauge block, employing the minimum number of task-specific projections required for surface-data completeness substantially outperformed a conventional high-projection scan. For a geometrically more complex test object, surface-related errors were still reduced within the region of interest. The proposed approach – particularly suited for flat surface structures and not accounting for image-degrading factors other than cone-beam artifacts – shows promise for high-throughput dimensional metrology of mono-material parts.