Optimization-based Structured Light Profilometry for Surface Reconstruction of Moving Objects

This paper presents a novel approach for accurate surface reconstruction of uniformly moving rigid samples from multi-shot structured light profilometry (SLP). Conventional triangulation-based multi-shot SLP cannot handle moving objects as motion during the acquisition phase causes errors in the assignment of pixel correspondence between cameras and projector. The proposed method utilizes an optimization-based, iterative strategy that considers the movement of the sample between consecutive captures to reconstruct the surface geometry. The optimization process is designed to minimize an objective function which evaluates the pixel value similarities between reprojected surface points in both the camera and projector images. The surface is reconstructed by iteratively refining the minimum search in the objective function using image filtering methods and gradient-based solvers. The method enables robust surface reconstruction of moving samples with a high measurement accuracy of 17 mu m for various directions and extents of motion, which can compete with static measurement results from conventional, high-accuracy SLP methods while outperforming them by a factor of more than 20 for large sample displacements, making multi-shot SLP accessible for industrial in-line measurement applications.