Three-dimensional scanning is a topic which, in recent years, has gained significant attention from several research fields, including industrial design, reverse engineering and prototyping (e.g., CAD), body imaging (e.g., fashion and health care), computer vision ( e.g., world mapping), entertainment (e.g., gaming, visual effects and animation), photo tourism, forensics and documentation of cultural artifacts. Currently, 3D scanning is an involved process that generally requires special hardware and procedures, and is therefore frequently limited to a laboratory environment. However, there is a world full of potential subjects to digitize, most of which cannot be relocated to a laboratory without considerable expense, or at all.
In this thesis, we introduce a 3D scanning system suitable for scanning objects and architecture using a mobile stereoscopic camera combined with an inertial navigation sensor. The inertial navigation system acts as a coarse estimator for rapid registration and camera localization. Our scanning apparatus is assembled entirely using commercially available hardware, and the scanning process itself requires no technical skill and our system may be adapted to a wide variety of scanning conditions. We also introduce a set of heuristics for multi view data registration that aim to produce reliable and repeatable convergence despite the diversity of potential subjects, and different types of motions that may be executed by the operator during the scanning process. A surface reconstruction technique is used to generate a mesh, and subsequent texturing of the model is done through interpolation of the per vertex colour information extracted from the rectified colour images recorded by the camera. By texturing the surface of an object in a 3D application, the realism of the product is greatly increased. The results presented will show that our system is a viable solution for reproducing fully textured polygonal models with low surface deviation and realistic visual appearance.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/28051 |
| Date | January 2009 |
| Creators | Byczkowski, Tomasz |
| Publisher | University of Ottawa (Canada) |
| Source Sets | Université d’Ottawa |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | 131 p. |
Page generated in 0.0111 seconds