Search

• Multilevel symmetric pattern is proposed for high-speed and high-accuracy 3D shape measurements. • The proposed method can achieve both speed and accuracy. • This method results in much better phase quality compared with conventional methods. • High-speed 3D shape measurement is performed at 667 Hz speed. The binary defocusing technique has enabled speed breakthroughs for 3D shape measurement, yet simultaneously achieving high accuracy and high speed remains difficult. To overcome this limitation, we propose to utilize multilevel symmetric pattern for high-speed and high-accuracy 3D shape measurement. Compared to conventional binary patterns, multilevel pattern could bring more flexibility for eliminating undesired high-frequency harmonics, thus has the potential to greatly enhance the phase quality and measurement accuracy. In this paper, the symmetric pattern design principle and related optimization procedure are presented in order to find the "best" multilevel fringe patterns. Both simulation and experiments verify that with respect to conventional methods, the proposed method could consistently generate better fringe patterns for a wide range of fringe periods. Furthermore, we developed an absolute 3D shape measurement system with the speed of 667 Hz, verifying that the proposed method is applicable for high-speed, high-accuracy applications. [ABSTRACT FROM AUTHOR]

Reducing the number of structured patterns for three-dimensional (3-D) reconstruction is of great importance for high-speed 3-D shape measurement. We present a method that reconstructs absolute 3-D shape using three projected binary patterns: one direct current (DC), one low-frequency, and one high-frequency fringe pattern. The procedures are (1) take the difference between the sinusoidal fringe patterns and the DC pattern; (2) apply Hilbert transform to the difference images to generate two phase maps; (3) employ the geometric-constraint-based phase unwrapping method to unwrap the low-frequency phase map; (4) unwrap the high-frequency phase map using the unwrapped low-frequency phase map; and (5) reconstruct 3-D shape. We developed a prototype system that can capture two-dimensional images at 6000 Hz, achieving 2000 Hz 3-D shape measurement speed. [ABSTRACT FROM AUTHOR]

In structured light 3D imaging, achieving robust and accurate 3D reconstruction with a limited number of fringe patterns remains a challenge. In this study, we introduce SFNet, a symmetric fusion network that designed for high-speed, high-quality 3D surface measurement using just two fringe images. The SFNet employs separate encoders and decoders for each fringe input to estimate its phase. The two generated phase values are then utilized to reconstruct the 3D information. During the training process, we use a refined reference phase which utilizes fringe images with different frequencies. SFNet has the capability to complement the additional frequency information by fusing the feature maps extracted from each encoder. Comparative experiments and ablation studies validate the effectiveness of our proposed method. The dataset is publicly accessible on our project page https://wonhoe-kim.github.io/SFNet/. [Display omitted] • We propose a 3D reconstruction framework utilizing deep learning based on Fringe Projection Profilometry. • SFNet is designed for phase retrieval only using two images. • The proposed method achieves accurate predictions of absolute phase with high precision. • We present a large-scale dataset for FPP technique within a virtual environment. [ABSTRACT FROM AUTHOR]