• Laser & Optoelectronics Progress
  • Vol. 58, Issue 20, 2010003 (2021)
Zichao Zhang, Zonghua Zhang*, Nan Gao, and Zhaozong Meng
Author Affiliations
  • School of Mechanical Engineering, Hebei University of Technology, Tianjin 300130, China
  • show less
    DOI: 10.3788/LOP202158.2010003 Cite this Article Set citation alerts
    Zichao Zhang, Zonghua Zhang, Nan Gao, Zhaozong Meng. U-Net-based Structured Light Three-dimensional Measurement Technology[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2010003 Copy Citation Text show less
    Procedure of neural network 3D measurement
    Fig. 1. Procedure of neural network 3D measurement
    Convolutional neural network structure
    Fig. 2. Convolutional neural network structure
    Part data in the dataset
    Fig. 3. Part data in the dataset
    Loss curves of the training set and validation set. (a) Loss curve of the training set; (b) loss curve of the validation set
    Fig. 4. Loss curves of the training set and validation set. (a) Loss curve of the training set; (b) loss curve of the validation set
    Experiment results of simulation. (a)(f)(k) Depth images; (b)(g)(l) depth images predicted by proposed method; (c)(h)(m) error images of proposed method; (d)(i)(n) depth images predicted by method in Ref.[20]; (e)(j)(o) error images of method in Ref.[20]
    Fig. 5. Experiment results of simulation. (a)(f)(k) Depth images; (b)(g)(l) depth images predicted by proposed method; (c)(h)(m) error images of proposed method; (d)(i)(n) depth images predicted by method in Ref.[20]; (e)(j)(o) error images of method in Ref.[20]
    3D effect. (a)(d)(g) 3D shape images; (b)(e)(h) predicted 3D shape images; (c)(f)(i) error images
    Fig. 6. 3D effect. (a)(d)(g) 3D shape images; (b)(e)(h) predicted 3D shape images; (c)(f)(i) error images
    Photograph of 3D measurement system
    Fig. 7. Photograph of 3D measurement system
    Deformed fringe patterns obtained by camera. (a) Deformed fringe pattern of mask; (b) deformed fringe pattern of human hand
    Fig. 8. Deformed fringe patterns obtained by camera. (a) Deformed fringe pattern of mask; (b) deformed fringe pattern of human hand
    3D shape data of real objects. (a) 3D shape data of mask; (b) 3D shape data of human hand; (c) detail display of mask eye; (d) detail display of human hand finger
    Fig. 9. 3D shape data of real objects. (a) 3D shape data of mask; (b) 3D shape data of human hand; (c) detail display of mask eye; (d) detail display of human hand finger
    Generalization capability analysis results. (a)(f) Depth images; (b)(g) depth images predicted by proposed method; (c)(h) error images of proposed method; (d)(i) depth images predicted by method in Ref.[20]; (e)(j) error images of method in Ref.[20]
    Fig. 10. Generalization capability analysis results. (a)(f) Depth images; (b)(g) depth images predicted by proposed method; (c)(h) error images of proposed method; (d)(i) depth images predicted by method in Ref.[20]; (e)(j) error images of method in Ref.[20]
    3D effect. (a)--(c) 3D shape data, predicted 3D shape data, and error image of sample 1; (d)--(f) 3D shape data, predicted 3D shape data, and error image of sample 2
    Fig. 11. 3D effect. (a)--(c) 3D shape data, predicted 3D shape data, and error image of sample 1; (d)--(f) 3D shape data, predicted 3D shape data, and error image of sample 2
    Anti-noise capability analysis results. (a)(e)(i)(m) Deformed fringe images; (b)(f)(j)(n) depth images; (c)(g)(k)(o) predicted depth images; (d)(h)(l)(p) error images
    Fig. 12. Anti-noise capability analysis results. (a)(e)(i)(m) Deformed fringe images; (b)(f)(j)(n) depth images; (c)(g)(k)(o) predicted depth images; (d)(h)(l)(p) error images
    3D effect. (a)(d)(g)(i) 3D shape data; (b)(e)(h)(k) predicted 3D shape data; (c)(f)(i)(l) error images
    Fig. 13. 3D effect. (a)(d)(g)(i) 3D shape data; (b)(e)(h)(k) predicted 3D shape data; (c)(f)(i)(l) error images
    Sample No.RMSE /%SSIM
    99200.650.9916
    100200.640.9921
    101200.880.9886
    Table 1. Error analysis of simulation experiment
    ObjectRMSE /%SSIM
    Mask1.420.9755
    Human hand2.250.9353
    Table 2. Error analysis of real objects
    ObjectRMSE /%SSIM
    Sample 10.940.9782
    Sample 20.890.9813
    Table 3. Error analysis of 3D object sample
    Noise levelRMSE /%SSIM
    0.5%0.610.9937
    1.5%0.630.9923
    2.5%0.700.9886
    3.5%0.780.9862
    4.5%0.870.9824
    Table 4. Noise analysis of validation set and test set
    Zichao Zhang, Zonghua Zhang, Nan Gao, Zhaozong Meng. U-Net-based Structured Light Three-dimensional Measurement Technology[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2010003
    Download Citation