Adaptive highlight suppression 3D reconstruction method based on binary-coded fringe patterns.

Abstract

When structured light-based 3D measurement is applied to highly reflective objects, specular highlights often cause saturation in bright regions, while low signal-to-noise ratios in dark regions lead to missing fringe information, thereby reducing measurement accuracy. To address this issue, this paper proposes an adaptive highlight suppression 3D reconstruction method based on binary-coded fringe patterns, which combines the robustness of binary fringes against gamma nonlinearity with the subpixel measurement accuracy of sinusoidal fringes. Specifically, the sinusoidal phase-shifting fringes are binary encoded, and an adaptive thresholding algorithm integrating a grayscale deviation coefficient with the grayscale mean is designed for binarization, thereby overcoming the problems of over-segmentation and under-segmentation in the binarization process. The binarized fringes are then decoded to achieve accurate recovery of the sinusoidal fringes. In addition, a camera imaging model is established to reveal the linear relationship between the binarization threshold and the exposure time. Based on iterative threshold calculation, an optimal exposure-time sequence is obtained, requiring only two projections of six images in total to complete exposure selection. On this basis, rapid HDR fusion is achieved using fringe modulation, generating high-quality HDR images and ultimately enabling subpixel-level 3D reconstruction. Experimental results demonstrate that the decoded sinusoidal fringes effectively improve the dynamic range of the fringe patterns. In the measurement of a metal flange, the proposed method achieves an average deviation of only 0.0025 mm and a standard deviation of 0.1478 mm. Compared with the EF method, the proposed approach requires only 25% of the number of exposures and improves point-cloud completeness by 34%. Further experiments verify that the proposed method performs well for highly reflective objects with surfaces such as metal and ceramic, exhibits strong applicability across systems with different hardware configurations, and can effectively achieve highlight suppression and accurate 3D measurement of highly reflective surfaces.