Deep learning-based fringe modulation-enhancing method for accurate fringe projection profilometry.

Area of Science:

• Optical Metrology
• 3-D Imaging
• Computer Vision

Background:

• Fringe projection profilometry (FPP) is a key 3-D measurement technique.
• Low fringe modulation, caused by surface reflectivity variations, leads to phase errors and inaccurate 3-D measurements.
• Existing methods for phase error reduction compromise 3-D shape details or measurement speed.

Purpose of the Study:

• To develop a novel deep learning-based method to enhance fringe modulation in FPP.
• To address the challenge of phase errors in FPP caused by low-modulation fringes.
• To achieve accurate 3-D FPP measurements without sacrificing speed or detail.

Main Methods:

• A deep learning-based fringe modulation-enhancing method (FMEM) was proposed.
• FMEM transforms two low-modulation, phase-shifted fringes into three high-modulation fringes.
• The enhanced fringes are used to calculate accurate phase data for 3-D reconstruction.

Main Results:

• The proposed FMEM effectively enhances fringe modulation from low-modulation inputs.
• Accurate 3-D measurements were achieved using the high-modulation fringes generated by FMEM.
• The method demonstrated effectiveness and accuracy in experimental analyses.

Conclusions:

• The deep learning-based FMEM successfully overcomes limitations of traditional FPP methods.
• FMEM provides a high-speed, high-accuracy solution for 3-D measurements on surfaces with varying reflectivity.
• This approach enables detailed and precise 3-D profilometry even with challenging surface conditions.