Eric Sauda, Emily Lancaster-Vine, and Natalie Morgan propose to apply a combination of traditional and cutting edge methods to model and visualize an urban area. We will model the entire collection of buildings within the urban, placing them accurately with respect to an underlying map or terrain representation. The initial urban area selected will by Uptown Charlotte, roughly defined as the area within the I-277 loop. This is an area that contains several skyscrapers, large hotels, government buildings, and other prominent buildings. In addition, it contains apartment buildings, condominiums, and houses.

We will provide 3D models for all these structures.  Our approach will employ data from multiple sources to construct a seamless, integrated 3D model of the urban area. Detailed 3D models of landmark buildings will be developed by hand using standard 3D modeling software such as MAYA. These models will be derived from a combination of oblique photography, digital plans (where available), and LIDAR data (to give precise rooftops). Care will be taken to develop consistent models that look good from all angles and that are appropriate for interactive, view-dependent visualization. For the latter, levels of detail will also be provided (for a given viewpoint, simple detail for a building in the distance and high detail for close-by buildings).

Click on the building below to see a Quicktime video of it's flythrough.

An additional set of buildings will be modeled by hand at lower levels of detail.  In discussion with our sponsors, we will determine the level of detail to be attained for these by hand models. More detail requires more modeling time, and, in addition, a highly detailed façade may require a different source of data (e.g., higher resolution photos and more of them) than its lower detail counterpart. We could plan, for example, 3 categories of buildings: high detail, intermediate detail, and lower detail. Important or landmark buildings (e.g., the stadium or the BoA tower) would be in the high detail category. The goal here will be to find a balance between effort and the visual quality of the urban model. This balance will be made easier to attain through the use of automatically generated semi-generic building models, as described next.  The rest of the buildings in the environment will be automatically modeled from insurance databases, which contain footprints and positioning (and sometimes heights, either directly or indirectly through the number of floors) for all buildings in an area. These automatically generated buildings will have generic façade textures, varied by building type (e.g., house versus small commercial building). We have found that these automatically generated buildings have accurate footprints and density and accurately follow the contours of the streets and neighborhoods. They thus provide a rich and useful context. Where desired, more accurate textures and roofs can be applied to any of these models.  

Using a set of novel automated methods based on urban legibility, we will generate progressive levels of detail for streetscapes, then neighborhoods, and whole districts. These levels of detail for aggregated buildings and associated structures will also have aggregated façade textures and skylines at varying detail. As collections of buildings grow to citywide sizes with tens to hundreds of thousands of buildings this will be the only way to maintain interactive rendering. With a separate interactive visualization capability that we have developed, the user will be able to fly anywhere and get a reasonable view of the city from any viewpoint. One could even add an extra capability for selecting and viewing levels of detail for very detailed building models so that the user can obtain sidewalk walk-by views. Thus with the modeling, levels of detail, and interactive visualization methods, we have the makings of a complete interactive city-viewing system that permits continuous navigation from overviews where most or the entire city is in sight to close-up views on individual buildings.