Fig. 1. Framework of the binary focused projection measurement method driven by deep learning

Fig. 2. Generating adversarial mechanism and network structure

Fig. 3. Branch residual Unet structure

Fig. 4. Measurement system and measured objects. (a) Fringe projection measurement system; (b) measured wood board; (c) measured ceramic vase

Fig. 5. Fringe pattern and wrapped phase. (a) Captured binary fringe pattern; (b) one period of wapped phase cross sections obtained by different methods; (c) wrapped phase calculated by binary patterns; (d) wrapped phase calculated by Unet^[17]; (e) wrapped phase calculated by the proposed method

Fig. 6. Fringe order and unwrapped phase. (a)‒(c) Fringe orders obtained by Graycode, ResUnet, and proposed method respectively; (d) unwrapped phase obtained by binary fringe+Graycode; (e) unwrapped phase obtained by Unet+ResUnet; (f) unwrapped phase obtained by proposed method; (g) cross-section of fringe orders; (h) cross-section of unwrapped phases

Fig. 7. 3D measurement results. (a) Binary defocusing+Graycode; (b) Unet+ResUnet; (c) binary focusing+deep learning; (d) sinusoidal fringe+Graycode; (e)‒(g) comparison of cross-sections in suitable defocused range, slight defocused range, and focused range

Fig. 8. Captured patterns at different depths. (a) Focused distance; (b) quasi-focused distance; (c) slightly defocused distance; (d) properly defocused distance

Fig. 9. 3D reconstruction results. (a) Binary pattern; (b) Unet+ResUnet; (c) proposed method