Automated Robot-Based Computed Tomography Trajectory Optimization using Differential Evolution in 3D Radon Space

Abstract

Limited accessibility of the X-ray hardware manipulating robots stemming from collision elements and the restricted workspace of the robots as well as areas of significant X-ray absorption are inherent characteristics of robot-based computed tomography scanning in subregions of large structures. The manual definition of trajectories is resource-intensive and results in substantial user influence on the resulting data quality. Therefore, this work proposes a method for the automated calculation of optimized (partial) circular scan trajectories for robot-based computed tomography. Specifically, a differential evolution algorithm is used to find global parametrization optima by estimating the reconstruction quality of trajectories. This estimation is based on a quantitative sampling quality metric in 3D Radon space, which is introduced in this work. The proposed method is evaluated on a test body from a region of limited accessibility within the strut mount of a car body. The reconstruction results are compared to those obtained from nearly 1000 reference trajectories. The results demonstrate that the proposed technique automatically generates trajectories that surpass the global optimum in data completeness of all reference trajectories. This methodology thus enables the elimination of user influence in trajectory parametrization.