Fig. 1. 3D measurement of traditional digital fringe projection. (a) Fringe projection; (b) wrapped phase; (c) absolute phase; (d) depth

Fig. 2. Single-stage deep learning based single-frame fringe projection 3D measurement method. (a)Fringe projection; (b)UNet; (c)Depth

Fig. 3. Multi-stage deep learning based single-frame fringe projection 3D measurement method. (a) Fringe; (b) FPTNet; (c) s phase-shifted fringe with different frequencies; (d) absolute phase; (e) PDNet; (f) depth

Fig. 4. 3D measurement results of two methods for a simple morphologic object. (a) Fringe projection; (b) measurement result of Deeplab V3+; (c) measurement result of ERFNet; (d) measurement result of UNet; (e) measurement result of the proposed method; (f) ground truth; (g) measurement error of DeeplabV3+; (h) measurement error of ERFNet; (i) measurement error of UNet; (j) measurement error of the proposed method

Fig. 5. 3D measurement results of two methods for a complex morphologic object. (a) Fringe projection; (b) measurement result of Deeplab V3+; (c) measurement result of ERFNet; (d) measurement result of UNet; (e) measurement result of the proposed method; (f) ground truth; (g) measurement error of Deeplab V3+; (h) measurement error of ERFNet; (i) measurement error of UNet; (j) measurement error of the proposed method

Fig. 6. (a) Error of multi-stage deep learning based single-frame fringe projection 3D measurement method; (b) corresponding enlarged detail of the red box in (a); (c) corresponding enlarged detail of the green box in (a)

Table 1. Main modules and parameters of PDNet

Table 2. 3D measurement results of two methods

Table 3. Error of multi-stage deep learning based single-frame fringe projection 3D measurement method on C3D test set

Table 4. Accuracy of measuring the standard sphere by using calibration parameter and PDNet