Master of Engineering - Research
School of Electrical, Computer and Telecommunications Engineering
He, Xiaochen, Registration of data generated by digital fringe projection profilometry, Master of Engineering - Research thesis, School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, 2012. https://ro.uow.edu.au/theses/3580
This research implements a digital fringe projection profilometry (DFPP) system that can acquire surface information of the 3-D object from multiple views, and also develops a program module that can register images from those views in the same coordinate system for 3D modeling.
In this research, a Digital Fringe Projection Profilometry (DFPP) system is employed for data acquisition. The DFPP system employs phase shifting profilometry (PSP) for the phase extraction. The optical geometry of the system is established for the phase to 3-D coordinates of a point cloud with respect to the system coordinate system. The system parameters are estimated through the system calibration module, which works in accordance with the fringe projection and acquisition module, the fringe pattern analysis module and the registration module. These modules achieve the fringe pattern design and projection, fringe pattern capture, projector and camera cooperation, captured fringe analysis, phase map extraction and 3-D point cloud construction.
The algorithm of registration proposed in this research is an enhancement of the conventional Iterative Closest Point (ICP) method with several extensions. The distance function for the correspondence finding is modified to be the weighted linear combination of positions and Euclidean invariant features for improving the probability of convergence. In addition, outliers can be discarded through robust statistics and an adaptive threshold of weighted distances between corresponding point pairs. The modifications aim to realize registration of a couple of partially overlapped point clouds, enhance the possibility of convergence and reduce the iteration times. This research evaluates the convergence behaviour and accuracy of the proposed registration method with Zhang’s conventional ICP method. In the ideal case, the proposed method was found to achieve the goals. The registration module in accordance with the existing fringe pattern projection and analysis module can align range images acquired from multiple views and thus integrate a 3-D model.