| Advancements in three-dimensional acquisition systems and technologies have enabled the state-of-the-art digital capture of real world objects and events with unprecedented detail, affordability, and ease. With these advancements come new opportunities for the digital modeling, processing, and analysis of the physical world directly from 3D measurements. However, due to the nature of the 3D acquisition process the acquired data often suffers from a variety of artifacts. Specifically for 3D scan data, artifacts such as measurement noise and outliers, non-uniform sampling, and incomplete measurements are a common occurrence when acquiring real-world objects and scenes. The presence of such artifacts if not addressed can negatively impact the accuracy and performance of applications that must process this data type.;In this thesis, the challenges associated with the robust processing of acquired three-dimensional data are examined. Acquisition artifacts common to 3D scan-based capture systems are considered and robust methods are developed for key processing tasks that comprise the geometry processing pipeline. Specifically, this research considers the problem of missing data and how it impacts the core processing tasks associated with range-scan alignment, point cloud orientation, and surface reconstruction. For each task, this work examines the missing data problem and develops robust methods for mitigating its impact. |
