Doctor of Philosophy
School of Electrical, Computer and Telecommunications Engineering
Cao, Pu, Implementation of fringe projection profilometry using triangular fringe patterns and spatial shift estimation, Doctor of Philosophy thesis, School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, 2015. http://ro.uow.edu.au/theses/4442
As an enabling technology for non-contact three-dimensional (3D) profile measurement, fringe pattern profilometry (FPP) has attracted intensive research due to the advantages of high accuracy, high speed and simply in implementation. Popular FPP approaches retrieve the 3D profile information based on the detection of phase differences. Recently, an alternative FPP approach, referred to as spatial shift estimation (SSE), has been proposed, which retrieves the 3D profile information based on spatial shifts. In contrast to phase difference estimation approaches, SSE approaches are advantageous as the fringe patterns do not need to be sinusoidal, and accurate reconstruction can be obtained even when nonlinear distortions exist on the fringe patterns. However, efficient implementation of SSE approaches is still an issue. This thesis aims to implement the SSE approach for FPP. Firstly, a digital fringe projection (DFP) system is designed and implemented, which is used as an experimental platform for the work presented in this thesis. A few problems associated with the implementation are studied and solved, including elimination of noise and distortion in the fringe patterns. An improved inverse function-based shift estimation method is then proposed to improve the performance of SSE approaches, which greatly reduced the computation time. Based on analysis of the limitations of traditional sinusoidal fringe patterns, a triangular fringe pattern is proposed. Theoretical analysis is presented to evaluate the complexity of the proposed triangular fringe pattern based algorithms, and practical experiments are performed to prove the efficiency of this proposed fringe pattern. Secondly, the shift unwrapping problem associated with SSE is investigated. Based on a review of the phase unwrapping problem in phase difference estimation based FPP, a reliability-guided algorithm and a multiple-wavelength algorithm are proposed to unwrap the shift map. Experiments are presented to demonstrate the effectiveness of the methods proposed. Finally, a selected-frequency shift unwrapping algorithm is proposed. An improved method is then introduced that uses selected wavelengths instead of frequencies. Combining the multiple methods introduced in the thesis, a novel 3D shape measurement method based on the projection of triangular patterns of two selected wavelengths is proposed. This method successfully solves the shift unwrapping problem of complex object surfaces with significant step heights or multiple separate objects, with the efficiency also increased since fewer images are used.