Projection Pattern Precorrection Method Based on Projection Error Decoupling in Fringe Projection Profilometry

Abstract

The quality of sinusoidal fringe projection is one of the most critical factors affecting the accuracy of fringe projection profilometry (FPP). Existing projection error correction methods, however, fail to address the simultaneous occurrence and mutual influence of geometric distortion and grayscale inconsistency in the actual projected pattern. In order to tackle this issue, this article proposes a projection pattern precorrection method based on projection error decoupling. By using multiangle expanded Gray code (MEGC), this method decouples the projection error correction process into two independent correction processes. Specifically, for the projection distortion error, a per-pixel distortion error calibration method is introduced. This method improves the accuracy of phase coordinate calculation by designing an encoding method for multiangle floating-point coordinates and a decoding method for weighted fusion of multi coordinate values. It compensates for the distortion error pixel by pixel using a nonparametric form. A subregional nonlinear error calibration method is, furthermore, proposed to enhance the sinusoidal quality of the projected fringes and ensure grayscale consistency. Finally, the precorrected pattern is generated based on the two error calibration results which ensures distortion-free and brightness-accurate fringe pattern. Experimental results with various 3-D objects demonstrated that our proposed method, achieves higher and more stable measurement accuracy (average 0.0327 mm versus 0.0184 mm), compared with traditional precorrection methods that address projection errors separately.