Shape reconstruction and inspection using multi-planar X-ray images

Abstract

Computer-aided manufacturing is widely used to create products from computer-aided design (CAD) models. The different characteristics of each manufacturing process can lead to product defects. Therefore, it is necessary to inspect the shape and dimensions of a manufactured product against the original CAD model for quality assurance. While the conventional laser-based three-dimensional (3D) scanning allows reconstruction of the external shape of an object, X-ray imaging can capture the internal shape. Aided by digital radiographic techniques, X-ray images can be captured within a few milliseconds. In this study, we used multiple X-ray images of an object to reconstruct its geometry and topology. Each of the two-dimensional (2D) X-ray images was processed to detect object edges. Using the geometric calibration information of the X-ray imaging system, the 3D vertices of the object were reconstructed. The topology, edges connecting vertices, was reconstructed by re-projecting the vertices to the X-ray images. The reconstructed object shape was statistically analyzed to understand the effects of the number of views and the parameters used in the reconstruction process on the accuracy of the reconstructed geometries. For our target objects, images from more than seven different views were used to successfully reconstruct the full shape of the object.