|Robust fringe projection profilometry
|Lun, Daniel P. K. (EIE)
|Three-dimensional imaging -- Technique.
Image processing -- Digital techniques.
Hong Kong Polytechnic University -- Dissertations
|Department of Electronic and Information Engineering
|xi, 132 pages : color illustrations
|Fringe projection profilometry (FPP) is a popular optical three-dimensional (3D) scanning methodology, which allows real-time measurements of an object's 3D information. It has salient advantages of low cost, high resolution, fast data acquisition, and full-field measurement as compared to other existing 3D scanning approaches. This thesis focuses on the development of robust FPP methods for obtaining accurate 3D estimation of the objects from incomplete or degraded observations due to the interference in the working environment. In a practical FPP procedure, some parts of the fringe images can be masked by the highlights generated due to the reflection of the surrounding global illuminations. In this research, an iterative inpainting regularization algorithm is proposed to restore the missing fringe patterns. The new algorithm detects the highlight regions automatically using a Gaussian mixture model (GMM). The geometrical structure of the missing fringe pattern is then sketched and used as the initial guess to guide the iterative regularization process. The simulation and experimental results show that the proposed method can detect the highlight regions and inpaint the missing fringe pattern accurately. They show that the proposed approach outperforms the traditional approaches in both quantitative and qualitative evaluations. Traditional FPP methods have the ambiguity problem that only the wrapped phase information (confined to -π to π) of the fringe pattern can be measured, although the true phase information is required. Various phase unwrapping methods have thus been proposed for the FPP; however, most of them have problem if the captured fringe images have a complex scene, for instance, containing multiple and occluded objects, having different kinds of artifact, such as high noise level in dark regions and discontinuities in fringe pattern, etc. In this research, a new marker encoding and detection algorithm is proposed to assist the phase unwrapping procedure to solve the ambiguity problem. For the proposed algorithm, unique markers that encode the true phase information of the fringes are embedded into the fringe pattern. Using the proposed marker detection and period order estimation algorithm, the markers are first detected and the true phase information is estimated accurately using a two-dimensional dual tree complex wavelet transform analysis. Then this true phase information is used to facilitate the phase unwrapping algorithm at the later stage. As shown in the simulation and experimental results, the proposed scheme is robust in obtaining the correct 3D model of objects with fringe images of complex scene. Besides, the algorithm is simple that does not introduce a significant burden to the FPP process computationally.
The above marker encoding and detection algorithm encodes the true phase information based on the position of the markers in the fringe pattern. A natural question arises if there is other form of encoding method that can give an even better performance in terms of robustness. In this research, we propose another algorithm which embeds a set of textural code patterns into the fringe pattern. It encodes the true phase information based on the morphological structure of the textural code patterns. During the fringe analysis procedure, the code patterns are separated from the fringe pattern using a modified morphological component analysis method. They are then decoded using a discriminative dictionary which is learned to give the sparse representations of the code patterns. They are integrated to a multilevel quality guide phase unwrapping procedure to allow the phase unwrapping to be carried out in fringe images of complex scene efficiently. The experimental results show that the proposed algorithm is superior over the traditional approaches in terms of robustness. It is also computationally efficient as it requires only approximately 300ms when running on a normal personal computer in the Matlab environment.
|All rights reserved
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: