Novel algorithms for phase extraction from one single shearographic fringe pattern and the recovery of obscured data

Abstract

The invention of laser in 1960s creates a new realm for physicists and engineers. This new light source provides an ideal medium for coherent optics and incubates a series of new experimental techniques such as holographic interferometry and shearography. Since their invention, holographic interferometry and shearography have provided powerful experimental tools for the development of theoretical mechanics and applications in nondestructive testing. In its early stage, shearography mainly provides qualitative measurement since shearographic fringe order can only be determined in integer level. With the rapid progress of computer technology and the advent of phase shifting technique in the 1980s, shearography gradually became a powerful tool for quantitative measurement with wide applications in both laboratory and industry. Accurate and robust as it is, the phase shifting technique requires capturing three or more fringe patterns at each deformed stage for phase determination, thus it is not suitable for dynamic measurement such as vibration, impact and fast deformation. To overcome this shortcoming, techniques based on Fourier transform or Wavelet transform which involve only one image are investigated and developed. However, such techniques always require generating a high frequency carrier fringe, which pose much difficulty in experiment design. Thus there is a great desire for a technique to extract shearographic phase information from one single fringe pattern. In this thesis, a novel phase evaluation approach based on the idea of phase clustering has been developed to extract phase information from one single fringe pattern. Before the object begins to deform, four speckle images are captured and stored. After the object starts a non-repeatable deformation, only one single speckle image is captured for each deformed stage. With these five specklegrams, the background information, fringe modulation and initial random phase can be determined. Since the deformation induced phase change is a smooth phase map which owns the property of phase clustering, it is successfully extracted from one single deformed specklegram using the special clustering approach developed in this thesis. A key step in applying the clustering approach for phase extraction is to guarantee the quality of the four speckle images captured before the deformation. These four images should be accurately phase shifted to ensure that the background information and fringe modulation are accurately obtained. Thus a proper calibration procedure of the phase shifter is proposed in this thesis, which helps to extract the deformation phase correctly. Simulated results as well as a series of experiments conducted on continuously deforming specimens show the ability and robustness of the proposed clustering phase retrieval method. Comparison between the proposed method and standard four-step phase shifting method further confirms the accuracy of the proposed method. The clustering phase retrieval method, however, may not work as well as phase shifting under some extreme cases such as very high fringe density or very noisy environment. This is because that the clustering approach utilizes only information from one deformed specklegram, while the phase shifting method has abundant information available from four deformed images. The result is that the clustering approach can work out a proper phase map for most of applications, but with some minor errors such as fringe interconnection and phase residues in some small areas for the extremely terrible situations. The fringe interconnection or phase residues may also happen when applying phase shifting method in very hostile environment. These erroneous data, which is called obscured data, will affect the phase unwrapping process and make it difficult to obtain the final unwrapped phase map for quantitative measurement. To recover the obscured data, two effective methods have been proposed and investigated in this thesis. The first one is to correct the erroneous areas using a phase filtering method called phase reclustering filtering which is a nonlinear filter specially designed for wrapped phase filtering and performs better than normal filters. By use of the reclustering approach, most of the interconnected fringes are separated and phase residues are cleaned. Thus the wrapped phase map becomes workable with normal phase unwrapping algorithm. The other approach tends to cut out areas with erroneous wrapped phase information, and then unwrap the remaining correct wrapped phase. After that, the unavailable unwrapped phase in erroneous areas is reconstructed from the surrounding data by a computerized tomographic method developed in this thesis. Simulated examples as well as a series of experiments conducted under extreme situations are used to verify the proposed phase reclustering filtering technique and the computerized tomographic phase reconstruction method. The results show that both methods are helpful. Compared with current filtering methods and reconstruction method, both proposed techniques have better performance. The three methods developed in this thesis have together provided a complete solution for quantitative measurement of continuously deforming object. With further refinement and development, they may open a door for interferometric techniques to be widely applied for ultra high-speed measurement in both laboratory and industrial environment.

Date of Award	16 Feb 2009
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Michael HUNG (Supervisor)