Single frame profilometry with rapid phase demodulation on colour-coded fringes
Loading...
Date
2019-07-01
Authors
Yee, Cong Kai
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Digital fringe projection profilometry (DFPP) is a non-contact whole-field surface profiling technique. Being able to produce dynamic measurement with high accuracy at video framerates of up to 4000 Hz, this technique is particularly useful in the biomedical field, industrial inspections and cultural heritage preservation. One primary challenge is increasing the measurement speed to achieve higher throughput and higher detectable rates of change. In this work, the applicability of De Bruijn colour-coded sinusoidal fringe projection pattern in achieving single frame profilometry was studied along with phase errors that occur due to gamma nonlinearity and colour crosstalk. Simplification of geometric model by using inherent slanted projection angle in off-the-shelf projector was also studied in this work. Therefore, a corresponding fringe analysis algorithm was developed for De Bruijn colour-coded fringe pattern to circumvent the conventional phase unwrapping techniques which are unreliable and time-consuming. A phase error compensation algorithm and a simplified geometric model were also developed to reduce the phase errors and number of model affliated parameters respectively. Three experiments were carried out to: (i) verify the applicability of the algorithms used in the proposed single-frame profilometry system; (ii) verify the applicability of the proposed phase error compensation algorithm; and (iii) compare the result of the proposed profilometry against the Mitutoyo CRYSTA-Plus M Series 196 coordinate-measuring machine (CMM). The experimental results showed that the proposed profilometry was able to reconstruct objects successfully and consistently using only a single-frame fringe image. The phase error compensation algorithm was also proven to reduce phase errors at reference level from (1.08 ± 2.32) % to (0.73 ± 0.46) % with a 95% confidence level for 8 iterations on average, producing a visually smoother reconstructed surface. In the comparison against the CMM on the reconstruction of a fan blade curved surface, the mean profile-to-profile differences were consistently recorded as 2.63 mm on average and the confidence intervals of the differences at 95% confidence level were recorded at around ± 0.20 mm. The findings prove that the proposed concept is applicable and provides an alternative method for conventional fringe analysis techniques such as transform-based algorithms or phase-shifting algorithms to advance high-speed 3D profilometry.