Single-shot fringe projection profilometry (FPP) offers significant advantages for dynamic three-dimensional (3D) measurement. However, existing deep learning-based single-shot methods still struggle to simultaneously and accurately extract the wrapped phase and fringe order. Because the former is a continuous physical quantity while the latter is a discrete integer label, their joint estimation from a single periodic pattern is inherently ambiguous, especially for isolated objects lacking spatial continuity. To address this problem, a deep learning-enabled single-shot FPP method is proposed by integrating inner shifting-phase encoding with a physically constrained dual-network architecture. The fringe order is embedded into the phase-shifting component of a single high-frequency fringe pattern, and two lightweight convolutional neural networks are then employed: one regresses the numerator and denominator of the wrapped phase, while the other classifies the inner shifting-phase value to recover the fringe order. This physically consistent decoupled design avoids the optimization conflicts of end-to-end regression and enables unambiguous absolute phase recovery. Experiments show that the proposed method could achieve high-accuracy absolute phase recovery comparable to traditional multi-frame techniques for measurements of complex and isolated objects, while maintaining low computational complexity. The dynamic experimental results demonstrate its application potential for dynamic 3D measurement.
It is challenging to stably and rapidly achieve accurate absolute phase retrieval for isolated objects with a single-shot pattern in fringe projection profilometry (FPP). In this context, a single-shot multi-frequency absolute phase retrieval (SAPR-DL) method based on deep learning is proposed, which only needs to capture one fringe image to obtain the full-field precise absolute phase. Specifically, a low-frequency deformed fringe image is loaded into the trained one-to-two deep learning framework (DLFT) to predict unit-frequency and high-frequency deformed fringe images. Then, three fringe images with different frequencies are loaded into the trained deep learning phase retrieval framework (DLPR) to calculate the corresponding absolute phase. The experimental results prove that the proposed SAPR-DL method can obtain the three-dimensional (3D) shape measurement of multiple complex objects by collecting a single-shot fringe image, showing great prospects in advancing scientific and engineering applications.
Intensity saturation can induce phase error and, thus, measurement error in fringe projection profilometry. To reduce saturation-induced phase errors, a compensation method is developed. The mathematical model of saturation-induced phase errors is analyzed for N-step phase-shifting profilometry, and the phase error is approximately N-folder of the frequency of the projected fringe. Additional N-step phase-shifting fringe patterns with initial phase-shift π/N are projected for generating a complementary phase map. The final phase map is obtained by averaging the original phase map extracted from the original fringe patterns and the complementary phase map, and then the phase error can be canceled out. Both simulations and experiments demonstrated that the proposed method can substantially reduce the saturation-induced phase error and realize accurate measurements for a highly dynamic range of scenarios.
Fringe projection profilometry (FPP) technique is a widely used method in three-dimensional measurement and has broad applications in industrial measurement. This method typically requires projecting multiple phase-shifted fringe patterns to achieve high-precision imaging. However, in industrial contexts, the measured objects are often metallic components with high reflectivity, which would adversely affect the reconstruction quality of FPP. High Dynamic Range (HDR) technology can effectively address such issues; however, it requires an increased number of projected fringe patterns, which reduces measurement efficiency. This paper proposes a single-shot 3D measurement method that combines traditional HDR techniques with deep learning to measure objects with highly reflective surfaces. The method employs two network models: the Phase Recovery Network (PRNet) and the Coarse Phase Unwrapping Network (CPUNet). PRNet receives a single fringe image with saturated pixels and outputs the numerator and denominator terms for phase retrieval. CPUNet takes the wrapped phase data processed through an arctangent function from PRNet’s output and predicts the coarse unwrapped phase to assist in phase unwrapping, yielding high-precision phase results. Experiments show that this approach requires only one shiny fringe image as input and effectively restores phase errors caused by high reflectivity, achieving a high reconstruction accuracy for the complete 3D shape of the object.