Imaging on the Edge: Mapping Object Corners and Edges with Stereo X-ray Tomography

X-ray computed tomography is a powerful tool for volumetric imaging, where three-dimensional (3D) images are generated from a large number of individual X-ray projection images. Collecting the required number of low noise projection images is however time-consuming and so the technique is not currently applicable when spatial information needs to be collected with high temporal resolution, such as in the study of dynamic processes. In our previous work, inspired by stereo vision, we developed stereo X-ray imaging methods that operate with only two X-ray projection images. Previously we have shown how this allowed us to map point and line fiducial markers into 3D space at significantly faster temporal resolutions. In this paper, we make two further contributions. Firstly, instead of utilising internal fiducial markers, we demonstrate the applicability of the method to the 3D mapping of sharp object corners, a problem of interest in measuring the deformation of manufactured components under different loads. Furthermore, we demonstrate how the approach can be applied to real stereo X-ray data, even in settings where we do not have the annotated real training data that was required for the training of our previous Machine Learning approach. This is achieved by substituting the real data with a relatively simple synthetic training dataset designed to mimic key aspects of the real data.