Measurement Accuracy Of Iterative Reconstructions With Limited Xct Data For Dimensional Measurements

Iterative Reconstruction (IR) algorithms can reconstruct X-ray Computed Tomography (XCT) data with high image quality and accuracy with limited data, leading to reduced acquisition times. The effect of limited data by reducing the total number of projections, N p , on measurement accuracy during IR reconstruction with SIRT and CGLS algorithms is examined in this work. The experiment consists of 2 parts - a preliminary study with a standard multi-sphere object and a comprehensive study on a complex aluminium object (NinjaStar) with Geometric Dimensioning and Tolerancing (GD&T) features, which is more representative of real-world applications. To achieve a similar measurement accuracy between the two algorithms SIRT requires fewer projections but takes more time to reconstruct. However, SIRT can handle noise more effectively, minimising form errors versus CGLS. Conversely, CGLS requires more data to obtain a similar measurement accuracy, but it reconstructs much faster than SIRT.