Among non-destructive testing (NDT) techniques utilizing X-rays, computed tomography (CT) plays a critical role in both medical and industrial applications by providing
three-dimensional tomographic images with excellent spatial and density resolutio...
Among non-destructive testing (NDT) techniques utilizing X-rays, computed tomography (CT) plays a critical role in both medical and industrial applications by providing
three-dimensional tomographic images with excellent spatial and density resolution. However, conventional CT systems are limited by the fixed size of the gantry and the circular or helical scanning trajectories they support. Moreover, when projection data containing highly attenuating materials are directly used during image reconstruction, streak-shaped metal artifacts can arise across the entire image, hindering accurate analysis of regions of interest.
Such limitations make conventional CT systems unsuitable for inspecting modern advanced products, which are often composed of multiple metallic components intricately combined into large and complex assemblies. To address these challenges, research has been directed toward developing robotic CT systems. In these systems, an X-ray source and detector are mounted on robotic arms, providing high degrees of freedom to acquire projections from arbitrary three-dimensional trajectories. This approach enables the avoidance of projection paths prone to artifact generation, thereby improving image quality. This thesis focuses on three main objectives. First, a single-arm robotic CT system was developed to conduct preliminary studies prior to building a twin-arm system, with the goal of demonstrating that comparable results can be achieved. Second, experimental studies were carried out to verify that non-standard, robot-controlled scanning trajectories can effectively avoid scan positions that induce artifacts.
Finally, irregular projection data acquired from non-standard trajectories were reconstructed using both analytical and iterative algorithms, followed by performance evaluation to identify the most suitable reconstruction approach for arbitrary scan trajectories.