Robust structured light measurement for large freeform surfaces

Structured light has shown substantial potential on large freeform surfaces measurements. However, existing methods encounter challenges associated with inaccurate initial pose estimation and cumulative errors in large-scale reconstructions. This paper introduces a robust automated measurement framework to address these issues. Initially, a structured light scanning platform is constructed to scan the target surface and acquire local point clouds. Subsequently, an initial pose optimization method based on a correction model is proposed to ensure high-quality initial poses for the local point clouds. Finally, a global registration method combining an improved graph optimization algorithm with the ICP (Iterative Closest Point) algorithm is introduced, enabling high-accuracy multi-frame registration and global optimization. Extensive experimental validations have been conducted. Compared to traditional methods, our approach reduces the RMSE (Root Mean Squared Error) from 1.43 mm to 0.93 mm, translation error from 0.98 mm to 0.53 mm, and rotation error from 0.21 degrees to 0.13 degrees. We believe that this study could provide a promising direction for achieving highly robust and accuracy 3D measurement over large areas.