Industrial Computed Tomography

Computed Tomography is well established in daily clinical routine as it enables radiologists to examine organs, tissues and pathological changes inside the patient. Likewise to medical examinations in the industrial context inspections of manufactured products are mandatory. Present quality measurements like tactile systems can only evaluate the products surface, whereas quality assurance, reverse engineering and dimensional measurement are also interested in information from below the surface. Industrial computed tomography (ICT) provides the ability to visualize both internal and external structures and therefore, overcomes this drawback of tactile systems. Conditions and requirements for reconstruction algorithms in ICT however differ from their medical counterpart. Industrial products are usually insensitive to an increased radiation exposure from examinations with higher energies and consist of less complex structures with known material characteristics. Furthermore, the reconstruction result needs to be of very high precision in its dimensions, which demands for specialized algorithms.

Currently ICT-algorithms use algebraic reconstruction techniques in order to create a tomogram of high quality. Based on a threshold and the tomograms voxel intensities ICT evaluation algorithms calculate the objects material interfaces and store them as polygonal meshes. Due to high tomogram and detector resolutions in combination with ART material interface extraction take a long time. Furthermore the polygonal mesh accuracy is limited by the tomogram resolution. In cooperation with the YXLON International GmbH, the Institute of Medical Engineering develops new reconstruction algorithms which overcome these drawbacks. The main idea is to skip the tomogram and rather reconstruct the polygon meshes directly. Therefore the algorithm starts with a polygon mesh of an arbitrary form and morphs it based on the projection images iteratively. The algorithm only handles the vertices of the mesh, which number is small compared to the number of voxels in the tomogram, leading to acceleration in the reconstruction speed. Furthermore, the mesh accuracy is no longer determined by the tomogram resolution.