Opto-Electronic Advances, Volume. 5, Issue 5, 210021(2022)

Deep-learning-enabled dual-frequency composite fringe projection profilometry for single-shot absolute 3D shape measurement

Author Affiliations
• 1Smart Computational Imaging (SCI) Laboratory, Nanjing University of Science and Technology, Nanjing 210094, China
• 2Jiangsu Key Laboratory of Spectral Imaging & Intelligent Sense, Nanjing University of Science and Technology, Nanjing 210094, China
• show less

Single-shot high-speed 3D imaging is important for reconstructions of dynamic objects. For fringe projection profilometry (FPP), however, it is still challenging to recover accurate 3D shapes of isolated objects by a single fringe image. In this paper, we demonstrate that the deep neural networks can be trained to directly recover the absolute phase from a unique fringe image that involves spatially multiplexed fringe patterns of different frequencies. The extracted phase is free from spectrum-aliasing problem which is hard to avoid for traditional spatial-multiplexing methods. Experiments on both static and dynamic scenes show that the proposed approach is robust to object motion and can obtain high-quality 3D reconstructions of isolated objects within a single fringe image.

Introduction

The development of information technology has accelerated human life into the digital three-dimensional (3D) world. Among many 3D optical measurement technologies, fringe projection profilometry (FPP) stands out as one of the most promising 3D imaging methods due to its non-contact, high spatial resolution, high measurement accuracy, and good system flexibility1-5. Nowadays, FPP has been widely applied in intelligent manufacturing, cultural relic scanning, human-computer interaction and some other fields6-9. In some important applications, such as rapid reverse engineering and online quality control10, 11, it is essential to obtain high-quality 3D information in continuously changing dynamic scenes12-14. For FPP, the projector projects a series of fringe patterns onto the target object, and then the camera captures these images modulated and deformed by the object. With the captured fringe patterns, the phase information of the measured object can be extracted through the fringe analysis algorithms. The most popular fringe analysis approaches are the Fourier transform (FT) methods15-19 and the phase-shifting (PS) methods20, 21. The FT approaches can utilize only a single high-frequency fringe pattern, where the phase information is recovered by applying a properly designed band-pass filter, such as the Hanning window, to extract phase-related spectrum information in the frequency domain. However, spectrum aliasing may cause low phase quality around discontinuities and isolated regions of the phase map. Unlike the FT methods, the PS technologies usually require three or more PS fringe patterns in the time domain to retrieve the phase map. Such methods are quite robust to ambient illumination and varying surface reflectivity, and can achieve pixel-wise phase measurement with high resolution and accuracy. For both FT and PS algorithms, the retrieved phase distribution is mathematically wrapped to principle values of arctangent function ranging between $−π$ and $π$. Consequently, the phase value is wrapped whenever there is a $2π$ jump. To solve the phase ambiguity problem and establish a unique pixel correspondence between the camera and the projector to ensure correct 3D reconstruction, phase unwrapping must be carried out. One of the most commonly used phase unwrapping methods are the temporal phase unwrapping (TPU) algorithms22, which can obtain the absolute phase with the assistance of multi-frequency fringe images. However, such sacrifice of time resolution using a large number of images seriously decreases the 3D measurement efficiency of FPP. Therefore, in order to measure dynamic scenes, researchers usually reduce the fringe patterns required for phase unwrapping, thus to improve the efficiency of per 3D reconstruction23, 24. Ideally, the absolute depth is expected to be obtained by a single-shot fringe pattern.

Tools

Get Citation

Copy Citation Text

Yixuan Li, Jiaming Qian, Shijie Feng, Qian Chen, Chao Zuo. Deep-learning-enabled dual-frequency composite fringe projection profilometry for single-shot absolute 3D shape measurement[J]. Opto-Electronic Advances, 2022, 5(5): 210021

EndNote(RIS)BibTexPlain Text
Paper Information

Category: