Pixel-wise phase map fusion technique for high dynamic range 3D shape measurement

Abstract Digital fringe projection technologies have been extensively applied in 3D shape measurements. However, it is still limited to achieve accurate 3D reconstruction of high reflective surfaces. To circumvent this problem, we propose a complete framework to realize pixel-by-pixel phase post-compensation of overexposed points, including the construction of hardware system and the guidelines of algorithm design. First, a virtual binocular vision system is established. Second, based on the fundamentals of projective and spatial transformation, three robust algorithms with low computational complexity are devised to generate a fused phase map that overcomes the information loss due to overexposure. Furthermore, experiments are conducted to evaluate the performance of the proposed method. Qualitative experiment results show that the presented technique achieve a similar surface coverage compared with the other two mainstream methods, but is superior to them in terms of measurement efficiency (13 images vs. 145 images, 13 images vs. 32 images). Quantitative experimentation further verifies the feasibility of our method in precise 3D measurement, with sphere diameter errors of 0.0367 mm and 0.0361 mm, spherical centroid distance error of 0.0269 mm, and plane distance error of 0.0186 mm.