Improvise: Interactive modeling promoting visual interpretation of the sensed environment
Abstract
CAD modeling and rendering systems have the intrinsic ability to provide much more information for the expectation-based scene interpretation task than just the synthetic images they provide as their primary output. There is a wealth of topological and geometric knowledge imbedded in the data structures which typical CAD systems use for their internal solid object representations, but, unfortunately, currently available graphics systems make no effort to maintain this information during the rendering process so that it can be used for model-based inspection or image understanding applications. In this thesis, we focus on one of these typical internal CAD data structures: the winged-edge boundary representation (BRep) for 3-D polyhedral solid modeling. We show how it can be integrated with other hierarchical data structures for efficient wireframe display and interactive manipulation by the user during the model construction process. We then show how to use the ray tracing method to render these BRep objects efficiently by taking advantage of the wireframe visibility information. We employ ray tracing because it allows us to implement sophisticated lighting effects which make our rendered scenes appear more realistic. The complete lighting model, including light source definitions and surface material properties, is described, along with auxiliary data structures and algorithms which make surface visibility and lighting computations more efficient. Finally, we describe the integration of greyscale and range image segmentation algorithms which process the registered rendered images in order to predict the appearance of elements from the solid model, as well as predict artifact elements which will be evident to the respective sensors modeled by the rendering process. This augmented symbolic information forms a synthetic expectation which can be used by a scene interpretation process to interpret the actual observed (sensed) data.
Degree
Ph.D.
Advisors
Kak, Purdue University.
Subject Area
Electrical engineering|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.