Fluorescent light error suppression for high-speed phase-shifting profilometry based on deep learning

Abstract

In the recording process of phase-shifting profilometry, intensity fluctuation caused by uorescent light source instability may occur and then introduce a non-ignorable phase error. More importantly, the selection of sampling speed will also affect the value of the phase error, which even up to 0.12 rad. To suppress this problem, a deep learning-based fluorescent light error suppression (DLFLES) method is proposed to achieve high-precise measurement under fluorescent light. Experiments demonstrate that the shapes of the reconstructed 3-D images are more precise using the proposed method. Our research would promote the development of accurate 3-D measurement under the interference of external light sources by using deep learning.