Advances in 3D optical profilometry

• Main Author: Huang, Lei
• Other Authors: Anand Krishna Asundi
• Format: Theses and Dissertations
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Manufacturing::Metrology
• Online Access: https://hdl.handle.net/10356/52124
• Institution: Nanyang Technological University

The increased significance and need for three-dimensional (3D) shape measurement in various industries for 3D inspection, quality control, and reverse engineering, has promoted the interest in 3D metrology. Among the 3D shape measurement techniques, optical methods show great advantages by offering non-contact, fast and flexible measurement systems compared to tactile tools such as the ubiquitous Coordinate Measuring Machine (CMM). However, various challenges still need to be addressed as the needs for accuracy, robustness, speed, and reliability increase. Basically 3D optical metrology has the following components – a pattern (fringe) illumination module, interaction of the fringe pattern with the object under test, recording of the fringe pattern using a digital camera (imaging module), and finally processing of the recorded results followed by display of the results. There are three types of surfaces – diffuse, specular, and transparent surfaces – typically encountered in engineering objects. The illumination, sample and imaging modules are treated as an entire system and combined effects are considered in the processing and data reduction stages. However, the phase which is obtained from the fringe processing is related to the profile in different ways for each of the methods. Fringe processing is one of the key aspects in optical 3D metrology. In the derivation of phase shift equations, a sinusoidal profile is presumed. Practically this is not always true and a novel scheme to analyze and compensate phase error from non-sinusoidal waveforms is proposed. Phase error is compensated through an iterative process based on analysis of the theoretical phase error. Experimental results demonstrate the feasibility of the proposed compensation method in removal of the periodical phase error due to the non-sinusoidal waveform. In addition, a practical phase retrieval framework is proposed for multi-frequency phase shifting method to accurately calculate the absolute phase and automatically identify invalid phase points via a thresholding and criterion checking process. For dynamic application, only a single frame of data is available to retrieve the wrapped phase and then unwrap it. A comparison of the transform-based algorithms for single-frame phase retrieval is conducted in the presence of noise and non-sinusoidal waveform conditions with simulation and experiment. The methods of 2D Windowed Fourier Ridges and 2D Wavelet Transform are the two superior phase retrieval algorithms in fringe projection profilometry. Quality-guided phase unwrapping technique is reviewed with a comparison on quality definitions and guiding strategies. When unwrapping noisy phase, transform-based methods perform better but with increased computation time. But none of the quality maps always succeeds in handling phase discontinuities. For guiding strategies, the classical guiding strategy with a data structure of indexed interwoven linked list is fastest to achieve the most reasonable unwrapping results. If a small unwrapping error is tolerated, stack-chain guiding strategy can works even faster. Next, some specific aspects related to methods for different surface properties (reflectance) are studied. First, with a consideration of camera lens distortion, an advanced least squares calibration method is proposed for fringe projection profilometry. The proposed method involves the distortion of camera lens to estimate the system parameters. It is more accurate than the existing least squares calibration method which is verified with both simulation and experiment. Second, an improved least squares integration method is proposed to enhance the accuracy of surface reconstruction from gradient which is obtained from the phase in a reflection arrangement. Significant improvement in accuracy is verified by comparing with the traditional least squares integration method. The merits of the proposed method are accurate, fast, and able to handle large datasets. A fast fringe reflection method with a single shot of a two-directional fringe pattern is then proposed for measuring dynamic scene of a specular surface. The influence of surface reflective property is next discussed through a study on 3D shape measurement of partially diffuse and specular surfaces using both the fringe projection technique and fringe reflection technique. Fringe reflection technique is found more sensitive on the partially diffuse and specular surface in the experiment. The combination of both fringe projection and reflection techniques is a practical solution as well. At last, a method is proposed to analyze the fringe patterns of double-side reflection from the front and rear interfaces of a transparent surface to retrieve the fringe phase from either surface for further dimensional reconstruction.