基于高阶多项式的条纹投影标定方法

This paper targets calibration for phase-shifting profilometry (PSP), where the system must estimate a reliable mapping from image coordinates and absolute phase to 3D coordinates. Traditional PSP calibration can be sensitive to setup accuracy and time-consuming when high-order polynomial models are fitted directly.

The proposed method keeps the high-order polynomial model but solves the fitting problem with a sparse greedy randomized Kaczmarz strategy. The greedy selection step prioritizes informative equations, while the randomized Kaczmarz iteration keeps the solver efficient for larger calibration systems.

This work appears in Chinese Journal of Lasers, 52(10):1004002-1004002.

Algorithm principle

The calibration model links camera pixel coordinates, absolute phase, and 3D coordinates. By fitting this mapping with sparse greedy randomized Kaczmarz iterations, the method avoids the instability and cost of direct high-order least-squares fitting while preserving the reconstruction benefit of a richer polynomial model.

Visual material

The available PDF for this record does not expose reusable embedded figure images. This page is therefore kept as a text-first summary until the original figure assets are added.

Original figures to add later include the iteration-error curve, comparative calibration and reconstruction results, and the ceramic model reconstruction comparison between least-squares fitting and the proposed SGRK-based calibration.

Results and impact

The experiments reported in Chinese Journal of Lasers compare calibration speed and reconstruction accuracy against conventional fitting strategies. The result is positioned as a faster calibration path for structured-light 3D measurement systems where high-order polynomial models are useful but direct least-squares fitting becomes inefficient or unstable.