Interactive measurements of three-dimensional branching objects

Abstract

One of the methods to learn more about the shape of an object is to measure it. Measuring the object yields quantitative data which can subsequently be used to gain new knowledge about the object. When the same metrics are applied to a number of objects, it becomes possible to perform an objective quantitative comparison of these objects. Performing the measurements using an interactive measurement system makes it possible to obtain and analyze large amounts of accurate measurements. This makes a more thorough analysis possible. Branching objects are ubiquitous in nature. Examples include trees, vascular systems, neurons, bronchial tissue trees, snowflakes, river networks, sea sponges, and branching corals. The focus of this thesis is on measuring the shape of marine branching corals. While these marine creatures are relatively simple and small organisms, they form large underwater colonies with a wide variety of different shapes that are in part dependent on environmental factors. These colonies are the cornerstone of many marine environments. Accurate quantitative analysis of these shapes enables biologists to better understand the mechanisms which govern coral growth. In this thesis, marine branching corals are subjected to quantitative morphological analysis. For the first time, a fully three-dimensional analysis is performed on 3D Computed Tomography (CT) scans of coral specimens. From these images the 3D branching structure is extracted by computing the morphological skeleton; this is an intermediary representation that is used to locate the relevant features of the coral. The extraction is not without difficulties, as different extraction algorithms have different sensitivity to noise and various other artifacts that are present in the data. This is solved in part by quantitatively comparing different algorithms and selecting the most appropriate one, and in part through interactive editing of the data to remove some of the artifacts. The extracted skeleton is used to perform the actual measurements. Some of the metrics use only the skeleton, while others use the skeleton to perform measurements on the image, for example to measure the thickness. Three coral specimens were measured and compared in this manner; the results were consistent with previously observed trends for the species, which validates the methodology. The use of visualization and interaction for a number of purposes is researched. The coral is explored using interactive 3D visualizations. In addition, such visualizations are used to edit the data in order to remove artifacts which would affect the analysis. The results of the measurements are also explored in a highly interactive manner. Interactive selection in linked 2D statistic data plots and 3D visualizations is used, where selection in one view is restricts or highlights corresponding parts of the other view. Finally, the use of non-photorealistic rendering techniques is used to add uncertainty to 3D visualizations. Interacting with complex objects such as corals can be difficult and confusing to the user, as only visual information is relied upon. A more natural user interface also makes use of the sense of touch; this is achieved by using tangible input props and augmented reality. These props are created by using 3D printing technology to produce a tangible representation of the data. Instead of using traditional input devices such as a mouse or 3D pen, the user interacts with the coral by holding it in their hand and manipulating it just like a real object. Augmented reality techniques are then used to annotate this hand-held prop with relevant visualizations, enabling efficient interaction with the real-world object. The methods and techniques in this thesis allow coral researchers to perform accurate measurements on their specimens. However, most of the concepts and techniques described in this thesis are not limited to coral. Skeletons are used to quantify and analyze a wide variety of objects. Interactive visualizations are also applicable to any field. Finally, printed tangible props can be created to interact with any type of data.