Objective  Fringe projection profilometry (FPP) has been established in a wide range of industrial applications. Its key advantages include high accuracy, fast speed, robustness, and non-contact operation. However, image saturation occurs inevitably in measuring high-reflective surface due to the light intensity exceeding the dynamic range of camera sensor. It leads to absolute phase errors and 3D reconstruction failures. It is not an effective approach to address this issue by roughly reducing the exposure time of the camera or shrinking the size of the lens aperture because the SNR of the fringe patterns would decrease. Therefore, measuring high-reflective surface is still a challenging task for conventional FPP. In this paper, we propose a block-smoothed adaptive fringe projection method for high-reflective surface measurement to achieve accurate 3D reconstruction.

  Methods  Firstly, the saturated region is extracted and then divided into several blocks by projecting several uniform gray patterns. The initial projection intensity of each block is calculated according to the saturation degree. Secondly, the surface is illuminated by one set of bright fringe patterns and one set of dark ones in the unsaturated condition. The absolute phase map in the extracted saturated region is obtained by fusing the phase maps of the bright and dark fringe patterns. The coordinate mapping between the camera pixels and the projector units is carried out by the integrated phase map. And the initial mapping projection intensity on the projector is gathered. Thirdly, the polynomial function is fitted to the initial mapping projection intensity. The function is utilized to construct the smooth projection intensity curve. The optimal projection intensity of each saturated pixel is refined. And the mapping holes caused by the inconsistent resolution of the camera and projector are filled simultaneously. The adaptive fringe patterns are eventually generated. Finally, the adaptive fringe patterns are projected to the high-reflective surfaces to implement accurate phase retrieval and three-dimensional reconstruction.

  Results and Discussions   Two metal parts and one porcelain were measured to verify the performance of the proposed method. Compared with the traditional and two previous methods, the proposed method is capable of generating more accurate and smooth optimal projection intensity (Fig.11-13). As a result, the image saturation is efficiently avoided while the SNR is guaranteed. The reconstruction of the proposed method achieved more smooth boundaries between the high-reflective and normal regions. Moreover, the reconstruction of the high-reflective region had no holes and less burrs. One metal plane was employed to evaluate the measurement precision of the methods (Fig.14). The measurement error of the proposed method decreased by 40% and 28.6% over the previous methods (Tab.1).

  Conclusions  The experimental results proved that the proposed method could generate more accurate and smooth optimal projection intensity map to deal with the image saturation problem while measuring high-reflective surface. The ambiguity of the refined optimal projection intensity is also corrected by using the proposed method, thereby establishing more accurate coordinate mapping between the camera pixels and projector units. In addition, the mapping holes are filled. The method improves the precision of measuring high-reflective surface effectively.