Reference free method for robot CT imaging geometry estimation

Abstract

X-ray computed tomography (XCT) is increasingly used for quality control and non-destructive inspection in the manufacturing industry. Compared to conventional XCT machines, twin robot CT systems are a promising platform for the scanning of oversized workpieces without disassembling them. This work proposes a reference-free geometry estimation method based solely on X-ray images of the specimen to compensate for the positioning and orientation errors of the X-ray the imaging system (i.e. X-ray source and detector). The identification of the system misalignment is based on the Grangeat-based consistency conditions, and is further optimized by a searching based method. We theoretically demonstrates that it is possible to recover the geometric parameters of a twin robot CT system based on solely the X-ray images of a rigid object. Simulated and real X-ray scans are used to evaluate the proposed method both qualitatively in terms of reduction of artifacts in the reconstructed volume and quantitatively in terms of the accuracy of the calculated geometric parameters. Several optimizers are compared, and the proposed method significantly outperforms the others.