Accurate dynamic 3-D shape measurement based on the fringe pattern super-reconstruction technique

Fringe projection profilometry (FPP) has been widely used for three-dimensional (3-D) shape measurement. However, measuring the dynamic objects is always problematic due to the movement of objects during the inter-frame time gap. Many dynamic solutions have been developed, but they are fragile for practical applications. In this paper, we propose a new method for dynamic 3-D measurement with flexible implementation and high measurement accuracy, which utilizes the inverse relationship between camera sampling speed and sampling resolution and image super-reconstruction technique. The camera selects a low sampling resolution to capture fringe patterns, thus improving the sampling speed and alleviating the influence of object’s dynamic motion. However, insufficient resolution will affect the accuracy and details of the reconstructed 3-D shape. Inspired by the successes of computational imaging techniques, the resolution and details lost by the front-end camera can be restored by the back-end image processing. Therefore, we specifically designed a details restoration and super-reconstruction network (DRSRNet) to transform the captured low-resolution fringes into the desired high-quality and high-resolution fringes. The desired 3-D shapes are calculated from these transformed fringes combined with the correspondence system calibration parameters. The provided experiments verify that the proposed method can improve the 3-D profiling speed by 5 times while ensuring the profiling accuracy, and the motion-induced errors can be obviously eliminated compared with FPP using the camera with a high sampling resolution. Furthermore, the proposed method can achieve different 3-D profiling speed increases to flexibly suit different measurement requirements.