Self-Calibration of the Key Geometric Parameters in Computed Tomography System Using an Entropy-Based Method

Correction of the geometric parameters in an industrial Computed Tomography (CT) system is an unavoidable step in its imaging process to avoid the reconstruction artifacts induced by the deviations in the geometric positions of components. When calculating the geometric correction parameter in industrial CT, two main types of methods are adopted: one is the offline multiparameter calibration method, which requires a separate phantom and extra scanning; the other is the online calibration method, which does not require extra scanning and relies on the projected or reconstructed images only, with the advantage of good real-time performance. The current online calibration method is computationally intensive and has low calibration accuracy. To solve this problem, we first analyzed the seven geometric errors of industrial CT systems and determined the parameters with a large influence degree. Then, we realized the initial positioning of the rotary axis by using the projection mirror characteristic. Finally, we designed a method for searching with two-dimensional entropy as the objective function for the transverse coordinates of the main point, which has the most significant influence degree. We used the fan-beam reconstruction method to accelerate the speed of a single reconstruction and realize fast online calibration. The experimental results show that our method can quickly calibrate the most sensitive u-parameter in the reconstruction and can apply this parameter directly to the reconstruction after correcting the geometric parameters. The reconstruction effect is significantly better than the reconstruction results before correction.