• CAD
• point cloud
• photogrammetry
• mesh
• 3D modeling
• surface reconstruction
• computer vision
• mesh assessment

CAD surface reconstruction from complex point clouds through an automatic data processing has not yet been solved in the literature, especially when dealing with big clouds. Adding to that, the 3D freedom layout, incompleteness, noisy, not uniformity and variable radius section of the input data are the main variables linked to the complexity above mentioned. Handle all these variables together would have a tremendous impact for the digitalization of natural or human artifacts as industrial systems or products.
This thesis has been carried out at UH Manoa and examines the case of a point cloud characterized by the above features, result of a close-range photogrammetry problem applied to a coral’s photos. Then, four different methods for CAD surface reconstruction have been developed and compared proposing new quantitative assessment framework on the quality of them. Each mesh generator, that offers a different degree of constraints on the input point cloud, output mesh quality and cost, must be chosen based on the target application.