Three-dimensional (3D) measurement of metal surfaces is one of fundamental tasks for product life-cycle management of complex equipment, which is meaningful but challenging due to its optical characteristics of highly reflectivity. To reliably reconstruct 3D metal surfaces, commonly-used techniques heavily rely on multiple exposures for optimal fusion but do not fit to high-efficient monitoring. To alleviate this reliance, we propose a novel single-exposure method called deep sinusoidal prior (DSP) for damaged phase recovery of highly reflective surfaces. Specifically, the sinusoidal hypothesis is instilled into an untrained deep neural network (DNN) as two-stream information, in order to bypass the problem of brightness enhancement. Utilizing elaborately designed loss functions, this approach enables us to restore the accurate phase encoding by fitting the DNN to two-stream sinusoidal priors. Experimental results demonstrate the proposed DSP method has superior performances on damaged phase recovery requiring no training samples. For instance, measuring a standard workpiece, absolute errors of the DSP method have been decreased substantially (81.69% and 59.49%) compared with the direct measurement and achieved similar accuracy (0.0754 mm vs. 0.0744 mm) compared with the reference. Most strikingly, the proposed method, for the first time, demonstrates a new perspective of recovering the reliable phase from a degraded one itself, contributing to the superior generalization capability insensitive to fringe frequencies, imaging settings and variant scenes.

High dynamic range (HDR) 3D measurement is a meaningful but challenging problem. Recently, many deep-learning-based methods have been proposed for the HDR problem. However, due to learning redundant fringe intensity information, their networks are difficult to converge for data with complex surface reflectivity and various illumination conditions, resulting in non-robust performance. To address this problem, we propose a physics-based supervised learning method. By introducing the physical model for phase retrieval, we design a novel, to the best of our knowledge, sinusoidal-component-to-sinusoidal-component mapping paradigm. Consequently, the scale difference of fringe intensity in various illumination scenarios can be eliminated. Compared with conventional supervised-learning methods, our method can greatly promote the convergence of the network and the generalization ability, while compared with the recently proposed unsupervised-learning method, our method can recover complex surfaces with much more details. To better evaluate our method, we specially design the experiment by training the network merely using the metal objects and testing the performance using different diffuse sculptures, metal surfaces, and their hybrid scenes. Experiments for all the testing scenarios have high-quality phase recovery with an STD error of about 0.03 rad, which reveals the superior generalization ability for complex reflectivity and various illumination conditions. Furthermore, the zoom-in 3D plots of the sculpture verify its fidelity on recovering fine details.

Three-dimensional (3D) shape measurement serves an important role in many areas, and fringe projection profilometry (FPP) is a widely used 3D measurement technique due to its non-physical contact and high speed. The real measurement scenarios are often mixtures of specular and diffuse reflections, causing overexposed and underexposed areas to co-exist. Currently, utilizing FPP to simultaneously measure overexposed and underexposed areas remains a challenge. To solve this problem, we propose a mixed reflection model and what we believe to be a novel high dynamic range (HDR) 3D measurement method based on polarization and multispectrum co-modulation. In mixed reflection, the fringe images captured by the polarized color camera can be modulated to different intensities between different channels due to the co-modulation effect. By synthesizing all sub-images, high-modulation fringe images are formed and simultaneous reconstruction of overexposed and underexposed surfaces is finally achieved. Compared to conventional methods, the proposed method is more effective for measuring complex reflection situations, especially when objects with specular and diffuse surfaces simultaneously exist in the scene. And what we believe to be a novel no-registration-error calibration framework for multi-channel cameras has also been proposed, which both acquires a significant amount of information in the region with HDR problems and avoids the registration error due to the physical distances between different channels. Experiments were conducted to verify the effectiveness of the proposed method.

Crossed-grating phase-shifting profilometry (CGPSP) has great utility in three-dimensional shape measurement due to its ability to acquire horizontal and vertical phase maps in a single measurement. However, CGPSP is extremely sensitive to the non-linearity effect of a digital fringe projection system, which is not studied in depth yet. In this paper, a mathematical model is established to analyze the phase error caused by the non-linearity effect. Subsequently, two methods used to eliminate the non-linearity error are discussed in detail. To be specific, a double five-step algorithm based on the mathematical model is proposed to passively suppress the second non-linearity. Furthermore, a precoding gamma correction method based on probability distribution function is introduced to actively attenuate the non-linearity of the captured crossed fringe. The comparison results show that the active gamma correction method requires less fringe patterns and can more effectively reduce the non-linearity error compared with the passive method. Finally, employing CGPSP with gamma correction, a faster and reliable inverse pattern projection is realized with less fringe patterns.

Laser 3D measurement has gained widespread applications in industrial metrology . Still, it is usually limited by surfaces with high dynamic range (HDR) or the colorful surface texture of measured surfaces, such as metal and black industrial parts. Currently, conventional methods generally work with relatively strong-power laser intensities, which could potentially damage the sample or induce eye-safety concerns. For deep-learning-based methods, due to the different reflectivity of the measured surfaces, the HDR problem may require cumbersome adjustment of laser intensity in order to acquire enough training data. Even so, the problem of inaccurate ground truth may occur. To address these issues, this paper proposes the deep feature recovery (DFR) strategy to enhance low-light laser stripe images for achieving HDR 3D reconstruction with low cost, high robustness, and eye safety. To the best of our knowledge, this is the first attempt to tackle the challenge of high measurement costs associated with measuring HDR surfaces in laser 3D measurement. In learning the features of low-power laser images, the proposed strategy has a superior generalization ability and is insensitive to different low laser powers and variant surface reflectivity. To verify this point, we specially design the experiments by training the network merely using the diffusely reflective masks (DRM951) and testing the performance using diffusely reflective masks, metal surfaces, black industrial parts (contained in the constructed datasets DRO690, MO191, and BO107) and their hybrid scenes. Experimental results demonstrate that the proposed DFR strategy has good performances on robustness by testing different measurement scenes. For variously reflective surfaces, such as diffusely reflective surfaces, metal surfaces, and black parts surfaces, the reconstructed 3D shapes all have a similar quality to the reference method.

Three-dimensional (3-D) measurement for high-dynamic-range (HDR) surfaces is one of challenge issues in industry manufacturing. However, current HDR methods, including conventional methods and supervised-learning-based methods, generally make compromise in either the efficiency or the accuracy. To alleviate this compromise, we propose a generalized fringe enhancement method based on the untrained neural network (UNN) to achieve HDR measurement from only a single exposure (SE). There are two important contributions in our work. First, to the best of our knowledge, we propose the first generalized UNN-based framework to solve the underexposure–overexposure hybrid issue. Without pretraining on any dataset, our framework can simultaneously achieve fringe denoising in the underexposure issue, and fringe inpainting in the overexposure issue. Second, we propose a sine regularization term to improve the inpainting quality in overexposed areas. Unlike existing methods that merely inpaint the corrupted areas based on their reliable adjacent areas, we utilize the unique sinusoid character of fringe to constrain the inpainting. Consequently, the robustness of our method can be effectively enhanced. Experiments for poorly illuminated scenes, high-reflection scenes, and their hybrid scenes demonstrate the proposed single-exposure method can substantially eliminate the measurement error (0.0603 versus 4.1775 mm) compared with the direct measurement.