A primitive-based 3D reconstruction method for remanufacturing

Localization and quantification of the defective volume of a broken part are critical for repairing but technically non-trivial, especially when the original CAD model is not available. To address this issue, a new geometric reconstruction method is proposed in this research, which features a novel primitive-based surface fitting and modelling technique. This technique combines the StepRANSAC (step random sample consensus) surface fitting and SDM (squared distance minimization) refitting, and this combination shows advantages in precision and robustness compared with the traditional RANSAC method, because main direction confirmation, segmentation, outlier filter, and refitting are applied in point clouds. Then, the defective volume can be localized and extracted by registering the reconstructed model and the damaged model. A few cases are studied to validate the effectiveness of the proposed method.