SegMo: Volumetric Image Segmentation System for X-rays CT Scanning of Mechanical Assemblies

Abstract

Industrial X-ray CT scanning systems (XCT) have been used for obtaining 3D shapes of mechanical parts. In this study we are interested in scanning a whole assembly made of several parts made of different materials. It has several advantages such as reducing scanning labor compared to that needed for scanning each of the parts by skipping disassembly process and also by reducing scanning operations. However, such whole assembly scanning is not a common practice as it is difficult to extract the parts from the volumetric image of the whole assembly. This process is generally called segmentation. Currently, an operator needs to spend long time on manually segmenting the volumetric image. This paper proposes a new segmentation program SegMo. It is based on a cell decomposition of the volumetric image with Morse complex defined with a function of the norm of gradient of the CT values of the image. The Morse complex has cell boundaries including parts’ boundaries. By constructing LOD (level of details) hierarchy of the Morse complex, a user can select an appropriate level of the hierarchy where the boundary of the target part is well reproduced and thus can form a region of a part. A system based on the SegMo with GUI (graphical user interface) is developed on ZVoxer system of Zodiac corporation. It has been applied to various examples to show that the SegMo system can perform segmentation more quickly and with less effort than conventional methods. In this paper, we describe the algorithm of the SegMo and its experimental results.