Surface Extraction by Accurate Fitting of Primitive Shapes to X-Ray Computed Tomography Scan Data

In recent years, active research has been conducted on X-ray computed tomography (CT) scans for observing and analyzing the internal structures and defects of products. High-precision is often required for products with primitive shapes, such as X-ray rotating ellipsoidal focusing mirrors, which are used for high-resolution observations. However, the CT reconstruction algorithm can cause blurring and artifacts in the CT volume, thereby complicating the accurate determination of the surface position. To address this issue, we propose an algorithm for high-precision surface extraction from the CT volume of primitive shapes. To accurately determine the surface position on the CT volume influenced by partial volume effects, we introduce three methods for primitive fitting. The first method approximates the boundary between the mirror and air in each cross-section with a line. The second method uses a local cylindrical approximation for the boundary, whereas the third method locally fits the primitive shape to the mirror for each cross-section. By comparing the outcomes of the proposed algorithm with those of conventional surface extraction algorithms, we demonstrate the superior accuracy of our approach and discuss the characteristics of various methods. Overall, the algorithm contributes toward enhancing the accuracy of internal structure analysis and defect detection in industrial components, potentially reducing manufacturing errors and improving product quality.