Pose imagery and automated three-dimensional modeling of urban environments

摘要: Three-dimensional (3-D) modeling of urban environments has numerous applications, including virtual environments, planning, and physical simulation. Constructing 3-D models from photographs (images) is thus an important area research in computer vision, increasingly, graphics. However, despite many years research, a system that automatically recovers realistic remains elusive; most practical systems require significant human input. Unlike automatic algorithms, human-assisted are not scalable, both terms the number images processed complexity generated model. This thesis describes novel techniques to extract textured pose imagery, i.e., annotated with camera position orientation single global coordinate system. Physical instruments (e.g., surveying, Global Positioning System (GPS), inertial sensors, etc.) used provide accurate initial estimates proposed algorithms. As these perfect, I first describe two optimization refine using information present images: spherical mosaicing relative rotations between taken position, mosaic registration accurately locates mosaics Next, algorithm extracts vertical facades pose. The employs horizontal line segments detect likely facade orientations space-sweep technique. Textures robustly computed for by combining several median statistics. I results large image dataset (consisting about four thousand eighty-one positions) office complex. These were successful recovering all complex, as well ell neighboring facades. (Copies available exclusively MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)