Improved calibration and 3D reconstruction for micro fringe projection profilometry

Micro fringe projection profilometry (MFPP) is a crucial three-dimensional measurement technology widely used in industrial inspection. As the system measurement accuracy is limited by calibration and reconstruction errors, this paper proposes a joint calibration of telecentric camera and projector, and develops noise-robust 3D reconstruction, ultimately forming a high-precision MFPP for microscopic object measurement. Specifically, the optimal parameters for the telecentric camera model are determined by constructing a virtual 3D calibration target, which circumvents the limitations that plane-based calibration methods often produce local optimal solutions. In response to the presence of phase errors, a method employing global homography is proposed to refine the projector pixel coordinates of feature points, thus enhancing the calibration accuracy of the projector. Furthermore, we integrate epipolar constraints into the reconstruction process, which helps eliminate noise points and optimizes the reconstruction results by constructing epipolar deviation. Extensive experiments demonstrate that our MFPP system can achieve micron-level measurement accuracy.