Deep learning-based frequency-multiplexing composite-fringe projection profilometry technique for one-shot 3D shape measurement

Chen, Yifei; Kang, Jiehu; Feng, Luyuan; Yuan, Leiwen; Liang, Jian; Zhao, Zongyang; Wu, Bin

doi:10.1016/j.measurement.2024.114640

Cited by 2 publications

(2 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The reason should be caused by the limitations of FFT. We refer to the latest research report of Li et al and Chen et al [17,18], who introduced convolutional neural networks into the composite FPP, which enhanced the adaptability to the measured object with complex surface geometry. However, this method requires a process of sample learning and training.…”

Section: Experiments and Discussionmentioning

confidence: 99%

“…Cao et al 's 2 + 1 composite method and An et al 's multi-frequency composite method used filtering windows to remove background DC components, but the usage of filtering windows is difficult to accurately distinguish the fundamental frequency and zero frequency components in the spectrum overlap region. Li et al and Chen et al intruduced deep learning technology into the composite FPP method to achieve a relative higher robustness for measured object with complex surface geometry [17,18], but the sample training and learning process is needed.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

3D reconstruction with single-frame two-step phase-shift method based on orthogonal composite fringe pattern projection

Wang,

Zhang,

Jiang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

In order to realize single-frame three-dimensional (3D) reconstruction, a single-frame two-step phase-shift method based on orthogonal composite pattern projection is proposed to solve the problem that the traditional N-step phase-shift profilometry needs multiple projections for 3D reconstruction. The orthogonal composite pattern uses only two carrier channels to reduce the spectrum overlapping influence on the demodulation accuracy of carrier and modulated fringes. A two-dimensional variational mode decomposition (2DVMD) method is adopted to remove the background DC component of the sinusoidal fringe to overcome the mode overlap problem by controlling the size of the bandwidth. Thus, the two-step phase-shift method is applied to calculate the phases for 3D reconstruction. The experimental results show that, compared with the typical Fourier Transform Profilometry (FTP) method , 3-step composite method and 2+1 composite method, the 3D reconstruction accuracy of the proposed method is improved by 49.1% ,31.4% and 23.2% respectively according to Mean Absolute Error (MAE), and by 73.0%, 58.4% and 56.8% respectively according to Mean Squared Error (MSE) as the evaluation index. Finally, the dynamic 3D reconstruction experiment demonstrates the good adaptability of dynamic 3D reconstruction.

show abstract

Section: Experiments and Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%